Rowerr Insights / AI SEO
The 5A system for SEO, AEO, GEO and AFO, with a real blog post and two WooCommerce product pages rebuilt step by step. No SEO plugin, no AI plugin, no paid tools.
By Abdul Wahab Taqi · Rowerr · Updated · 42 min readQuick answer
To rank in Google and get recommended by ChatGPT in 2026, build pages with real value, not just content. Let AI bots in (Access), put the answer in the first 60 words (Answer), give products clean, specific facts (Attributes), show the real person and proof behind the page (Authority), and get others talking about you (Amplify). It’s one system for Google and AI, and you can do all of it for free.
Key takeaways
No, SEO is not dead. Google still sends far more traffic than all AI tools combined, and most AI answers are built on top of search results. What stopped working is the old tricks and content without value. A page with real facts, real numbers and real experience ranks in Google and gets cited by ChatGPT, Gemini, Claude, Perplexity and Google’s AI Overviews.
You have published hundreds of AI blog posts and still get no traffic. You followed every Yoast and Rank Math rule, got a green score everywhere, and the result is zero. This guide fixes both problems.
Here is the proof from my own website. One of my posts on rowerr.com has a Rank Math score of 94 out of 100, a perfect green score. It is not on the first page of Google, and it got exactly one search impression. Another post on the same website scores 92. That one is on the first page of Google, and Google’s AI Overview uses it as a source.
So what is the difference? Not the score and not the plugin. The difference is value.
Rule #1: content vs value
Content is words on a page, and AI now writes it in seconds. Value is a page that really answers the question, with real facts, real numbers and real experience. Content alone fails now. Value ranks: in search engines, in AI engines and in answer engines.
In Part 5, I give you the eight things AI checks before it recommends your content. Miss one, and AI simply recommends someone else.
I started working on websites in 2014. In the twelve years since, I have built websites and mobile apps and handled their SEO and their content. I have seen Google change its rules again and again, but the last two years brought the biggest change I have seen. The rules did not just change. They were redrawn, because of AI.
Everything here comes from my own experience: what worked for me and my clients, and what did not. I also went through more than fifteen hours of the best SEO and AI search courses on YouTube and read almost five hundred comments under those videos, so I know the questions people actually ask. Where I agree with those experts, I say so. Where I don’t, I say that as well.
Even Google’s own guide for AI search says it clearly: this work is still SEO. What is dead is the old tricks, and content without value. Here is the biggest of those tricks.
Most AI SEO courses skip e-commerce, and the comments under them keep asking the same thing: what should I do for my store? Nobody answers them, so let me answer clearly.
Suppose you and a competitor sell the same product, but theirs has a few extra features that yours doesn’t. The old technique was to mention those features in your product description without ever clearly saying whether your product has them. You wrote around them, so Google would see the keywords.
Today that trick doesn’t just fail. It actively hurts your website, your content and your product rankings. AI reads your page like a list of facts. If the page is not clear, AI cannot match your product with the customer’s question, so it recommends someone else. And if a customer buys because of a confusing description and the feature isn’t there, you get a return and a bad review, and that review becomes one more fact AI reads about you.
What works now: clear product information, clear use cases (who it is for and what it does), clean attributes with no confusion, a clear page structure and an obvious add-to-cart button. Not a confusing product page, and not a messy product feed.
You don’t need any SEO plugin, any AI plugin or any paid tool for this. Everything in this guide is free, and in Part 7 I prove it on a store that has no SEO plugin at all.
If you run a store or a startup, or you are starting a new business, you are sitting on a gold mine. Most of your competitors have not started this yet, and it works in any country and in any language, not only in the US.
I call the whole method the 5A system. Do these five things, and the same page works for Google and for AI. One system, not two.
AI bots must be able to read your website.
It sounds simple, but one default setting can silently block them. Part 7 shows you which one.
Every page answers the question in its first lines, not after three paragraphs of story.
Products get clean, specific facts: size, price, ingredients, who it is for, and also what it does not do.
Other people talk about you: Reddit, YouTube, reviews and articles.
AI trusts what others say about you more than what you say about yourself.
Here is the plan for the rest of this guide. First, why a new business cannot win with the old SEO tricks, and why AI is your shortcut. Then I rebuild a real blog post in front of you, then two real product pages. Then AI readiness, the plugin question and thirty tips. And at the end, a one-hour checklist you can start today.
Let’s start with the question every new business owner asks me: should I focus on Google, or on AI?
For a new business, AI search is the bigger opportunity. It is faster, the visitors convert better, and people trust what AI recommends. But AI search is built on top of SEO, so you don’t build two things. You build one page, the right way, and it works for both.
Suppose you start a new business today: a store, an app or a service. You build a nice website, write some blog posts and follow all the old SEO guides. And nothing happens.
The reason is history. Many of the websites at the top of Google have been there for ten or fifteen years. They built backlinks, authority and thousands of pages long before 2024, and Google trusts them because of that history. If you copy the old tricks, like stuffing keywords, writing three-thousand-word articles or buying links, you are playing their game on their ground, years behind. You will not win that game.
I am not the only one saying this. Ethan Smith runs Graphite, an SEO agency that works with companies like Webflow. When startups ask him for SEO help, his first answer is: don’t do it at all. Without authority, you cannot rank in Google for the big keywords.
“You can get mentioned by a citation tomorrow and start showing up immediately.”
That changes everything. AI doesn’t just show the number-one blue link. It reads many sources and builds one answer from all of them. So if a Reddit thread, a YouTube video or a blog mentions you, you can show up.
There is a funny story in the comments under his video. Someone asked which agency is best for AEO. Another viewer replied, “just ask ChatGPT.” And Ethan himself replied that, apparently, the answer is not Graphite. Even one of the top experts in this field was not showing up in ChatGPT for his own service. The game is new for everyone, and that is your opportunity.
Before, you typed six words into Google, opened ten or fifteen websites, read them all and then decided. Now people ask ChatGPT a full question, with their budget, their skin type and their city, and get the answer directly. One estimate puts the average AI prompt at around twenty-five words. Those very specific questions are exactly where a small business can win.
“What’s a good fragrance-free moisturizer for dry, sensitive skin under $40 that works under sunscreen?”
On speed: one founder wrote that Reddit got his new product into AI answers in about one month, while SEO took seven months for his previous product. If you want to know whether Google is noticing your new site, check the impressions in Search Console. Impressions move before clicks.
On visitor quality: Ahrefs, one of the biggest SEO tool companies, checked its own numbers. AI search was only half a percent of its traffic, but it brought twelve percent of its signups, and AI visitors converted twenty-three times better than Google visitors. Webflow saw six times better conversion from AI traffic. As one viewer put it, these leads arrive already sold.
This is where a lot of guides exaggerate. Google still sends far more traffic than AI: around two hundred times more than the top AI platforms combined, according to Ahrefs. For e-commerce, the gap is smaller. One study of ninety-four stores found AI visitors converted at about 1.8 percent, against 1.4 percent from Google. Better, but not magic.
So I am not telling you to ignore Google. AI search is built on top of SEO, and the page that wins in AI is the same page that ranks in Google.
Your superpower as a new business
Old websites carry baggage: outdated pages, old prices, an old phone number, three different descriptions of the same product. AI hates contradictions, and when it finds them, it often cites nobody. You start clean, with no baggage.
Enough theory. Next, a real blog post on the demo store I built for this guide. Honestly, it’s not good, so let’s fix it together. And in Part 4 you’ll see a mistake I made on my own product page that you are probably making too.
Use one main keyword in the title, slug and H1. Turn real customer questions into H2 headings. Add a real author, dates and key takeaways, then answer the question in the first sixty words. Start every section with its answer, add a comparison table, useful images and a short FAQ, and refresh the page every one to three months.
The example is LUMÉRA, a demo skincare store I built for this guide, so I can show every step without showing any client’s data. The post is “How to Choose a Moisturizer for Your Skin Type.”
Let’s check it honestly. It’s only about two hundred words. The first line says, “The best moisturizer is the one you enjoy using every day.” It’s a nice line, but it doesn’t answer the question. The headings say “If your skin feels oily or shiny” and “If your skin feels dry or tight”: friendly, but no customer types that into Google or ChatGPT. There is no author, no key takeaways, no table and no FAQ. Google might index this page, but AI will probably skip it.
The main keyword is “how to choose a moisturizer for your skin type.” Use the same words in the title, the URL slug and the H1 heading. The title Google shows should start with the keyword. The rest of the title can change, but the keyword stays at the front.
One page can rank for twelve to twenty related searches, not only one. Take those related searches and turn each one into a heading, written as the question a customer actually asks, such as “What moisturizer is best for dry skin?” or “Gel or cream: which one do I need?”
Where do these questions come from? From your real customers: support emails, WhatsApp messages, reviews, Reddit threads and Google’s “People also ask” box. You can also ask ChatGPT, “What would my customer ask about choosing a moisturizer?” It gives you the follow-up questions that AI asks itself before it answers.
One warning. The most-liked comment under Ethan Smith’s video pointed out that many AI visibility tools suggest questions based on your own website, not on what customers actually ask. Don’t use those tool lists as your base. Start from your real customers.
Two more rules. First, use the words your customers use, not expert words. A chemist says “occlusive”; a customer says “heavy cream.” Second, don’t create a separate page for every small question. Answer the follow-ups on the same page. Google’s own guide says that making pages for every variation of a question can break its spam policies.
And never use the same main keyword on two different pages. If two of your pages fight for the same search, Google gets confused and often ranks neither.
At the very top, add a real author whose name links to their LinkedIn. Skincare is a trust topic, so if you can, have a qualified person review the page and mention it. Then show the published date and the updated date, followed by three to five key takeaways, because people want the answer fast.
Then the most important part of the whole page: answer the question in the first sixty words.
“To choose a moisturizer, match the texture to how your skin feels by mid-afternoon: a light gel for oily skin, a lotion for combination skin, a rich cream with ceramides for dry skin, and a short, fragrance-free formula for sensitive skin. In winter, go one step richer.”
People decide in the first few lines whether they will stay or leave, and AI reads the top of a page much more than the bottom. One widely quoted figure says 44 percent of ChatGPT’s citations come from the first third of a page. If your answer is in paragraph four, the reader and the AI have both moved on.
Under each H2 question, the first one or two sentences answer it directly. The details, examples and exceptions come after. Under “What moisturizer is best for dry skin?” the first line is: “Dry skin does best with a richer cream that contains ceramides, shea butter or squalane, because they help hold water in the skin.”
Notice that I named real ingredients: ceramides, squalane, not “good ingredients.” Specific names are clear for the reader, and AI trusts specific answers.
Some people will tell you that you must cut your content into tiny pieces for AI. Google’s own guide says there is no need for that, and there is no perfect page length. Write clear sections for people. If a section makes sense when someone reads only that part, it’s good.
| Skin type | Best texture | Look for | Go easy on |
|---|---|---|---|
| Oily | Gel or water-gel | Niacinamide, hyaluronic acid | Heavy oils and butters |
| Combination | Light lotion | Hyaluronic acid | Very rich creams all over |
| Dry | Rich cream | Ceramides, shea butter, squalane | Alcohol-heavy formulas |
| Sensitive | Short, fragrance-free formula | Ceramides, colloidal oat | Fragrance, essential oils |
One SEO agency I studied puts the comparison table at number one on its AI checklist, and it makes sense. The reader can compare everything in one look, and so can AI. Sometimes Google even shows your table inside its AI answer.
Use three types of images: a 16:9 featured image for the top of the post and for search results, a simple diagram such as a skin-type chart, and real photos instead of stock photos whenever you can. Give every image a file name and alt text that say what is actually in it, and use the WebP format so the page loads fast.
At the bottom, add an FAQ section. The FAQ heading is an H2 and each question is an H3. Write the questions the way people type them into ChatGPT, like “Can I use the same moisturizer morning and night?” or “Should moisturizer go before or after sunscreen?”, and keep every answer to two or three sentences.
Then links. The dry-skin section links to the product that fits it, and the post links to the serum guide and the daily routine guide on the same website. Add three to five calls to action that match what the reader wants right now, like “Shop fragrance-free moisturizers”, not a random newsletter pop-up.
The post went from about 200 words to about 1,200 words, but not because longer is better. Ahrefs looked at more than 174,000 pages cited in AI Overviews and found basically no link between word count and getting cited. More than half of the cited pages were under 1,000 words. Our page got longer only because it now answers more questions.
Keep it fresh. Update it every one to three months with real changes, not just a new date. Pages that AI cites are usually fresher than normal search results.
A blog post brings people to your website, but a store makes its money on the product page, and that’s where most stores fail in AI search. Before we fix two product pages, you need four minutes on the new rules.
SEO helps search engines find and rank your page. AEO makes your content the direct answer. GEO gets your brand named inside AI-written answers. AFO makes your website easy for AI agents to browse and buy from. RAG is how AI answers: it searches, reads, writes and cites. Around 80 percent of the work is the same for all of them.
SEO (search engine optimization)
Making your pages easy for search engines to find, understand and rank. You compete for a position in a list of results, and the user clicks your link.
Answer engine
Anything that gives a direct answer instead of a list of links: Google’s AI Overviews, ChatGPT, Perplexity or a voice assistant like Alexa.
AI engine (generative engine)
An answer engine that runs on a large language model. It doesn’t copy one page; it reads many sources and writes a new answer from them.
AEO (answer engine optimization)
Making your content the answer.
GEO (generative engine optimization)
Getting your brand inside the answers that AI engines write.
AFO (agent-friendly optimization)
Making your website easy for AI agents that browse, click and even buy on behalf of a person.
SEO is an agreed name that everybody uses. GEO and AFO are not agreed standards. Nobody official created them; they are terms people made.
GEO came from a research paper, and then marketers started using it. AFO comes from Google, in a way. Google’s web team published a guide about agent-friendly websites, with tips like clear buttons, stable layouts and proper labels on forms, so AI agents can use your site. From “agent-friendly,” people created “agent-friendly optimization,” and some call it agent findability optimization. Google’s own AI search guide mentions AEO and GEO, and then says this work is still SEO.
So don’t buy a service just because of a new acronym. Ask what work they will actually do.
The idea behind AFO is real, especially for stores. Suppose an AI agent is shopping for your customer. It needs to find the product, read the price, select the size and click add to cart. If your page is confusing, the agent fails, and you lose the sale.
RAG means retrieval-augmented generation. AI knows things in two ways. The first is training data: a giant snapshot of the internet that the model learned from. It is old and gets updated maybe every few months, so if you launched your product last week, the AI’s memory doesn’t know you. The second is retrieval. When a question needs fresh or specific information, AI runs a search, reads the pages it finds, writes the answer and cites some of them. That is RAG.
AI doesn’t run just one search. It breaks your question into many small searches, which is called query fan-out. Ask “Which moisturizer should I buy for dry skin?” and behind the scenes AI might also search “best ingredients for dry skin”, “cream or gel” and “fragrance-free moisturizer under $40”.
Research quoted by Ahrefs says an average prompt fans out into around nine to eleven searches, and more than 95 percent of them show zero search volume in keyword tools. Nobody is tracking them, but if your page answers them, you get pulled into the answer. That’s why the blog post in Part 2 covered the whole topic, not one keyword.
The most important line in this part
Most AI answers are a summary of a search. If your page is not in the search results that AI reads, you are not in the answer. That’s why SEO is still the foundation.
“Should I use vitamin C serum in the morning or at night?”
One direct answer, quoted from the page that answered best.
You win by being the quoted answer.
“Recommend a fragrance-free barrier cream under $40 for dry, sensitive skin.”
A short list of brands, built from many sources.
You win by being one of the names, ideally first.
AEO is when someone asks a direct question, like “Should I use vitamin C serum in the morning or at night?” AI gives one answer and quotes the page that answered it best. Our serum page says it clearly: mornings, after cleansing, before moisturizer and SPF. Winning AEO means being the quoted answer.
GEO is when someone asks for a recommendation, like “Recommend a fragrance-free barrier cream under $40 for dry, sensitive skin.” AI builds a short list of brands from many sources. Winning GEO means being one of those names, ideally the first. That depends on two things: the facts on your own page, and how many other places on the internet mention you.
For a new business, AEO is mostly about your pages. GEO is your pages plus your reputation on the web.
| Google SEO (old rules) | AI search (new rules) | |
|---|---|---|
| You compete for | A position in a list of ten links | A mention inside one answer |
| You target | One keyword per page | A whole topic and its hidden follow-up questions |
| A typical search | A few words | A long question with context, then follow-ups |
| Who wins the big terms | The number-one result | The brand mentioned most across many sources |
| Authority comes from | Backlinks and domain authority | Brand mentions and consistent facts across the web; links still help |
| Length | Long articles often looked better | Length doesn’t matter; answers do |
| Freshness | Helpful | Strongly favored |
| A new site can show up in | Months to years | Days to weeks, for specific questions |
| Results | Fairly stable | Change from one run to the next |
| You measure with | Search Console rankings and clicks | Share of voice in AI answers, AI referral traffic and “How did you hear about us?” |
Two rows matter most. First, authority. Ahrefs studied 75,000 brands, and the factor most strongly linked to showing up in Google’s AI Overviews was brand mentions across the web, stronger than backlinks. Fewer than one in three AI mentions even include a link. Being talked about now matters as much as being linked to.
Second, the results change. Ask the same question five times, and you might be named three times. It’s a probability, not a ranking, so never judge your AI visibility from one screenshot.
And what stays the same? A fast, clear, honest and helpful page wins in both worlds.
Now you know the new rules. Let’s use them where the money is: on the product page. You’ll also see that mistake on my own product page.
Give the product a clear name and slug, make the short description answer what it is, who it’s for and the sizes, and show a real price for every variation. Answer the follow-up questions in the description, use global attributes with one clean value each, never type prices into text, say clearly what the product does not do, and keep your Merchant Center feed identical to the page.
Here are two products from the demo store: a variable product with two sizes and a simple product. Between them, they cover almost every situation a WooCommerce store has.
The two-second rule
Within two seconds of landing on the page, a customer should know three things: what the product is, how much it costs, and where the add-to-cart button is. If they have to search for these, the page is confusing. And if a human is confused, an AI agent is confused too.
“Vitamin C Radiance Serum”, with the slug vitamin-c-radiance-serum. No fancy names. The brand, LUMÉRA, is set in WooCommerce’s own Brand field instead of being stuffed into the title.
“A lightweight vitamin C serum with ferulic acid and orange peel extract for a fresh, refreshed-looking complexion. The bright first step after cleansing each morning. Available in 30 ml and 50 ml.”
In three short sentences you get what it is, the key ingredients, when to use it and the sizes. It’s the blog rule again: answer in the first sixty words. The short description is the first thing a human and a machine read.
There are two sizes, and each has its own SKU, price and stock: 30 ml at $42, down from $48, and 50 ml at $68. AI cannot recommend you for “under $50” if it cannot find your price.
“Why you’ll love it”, a product details list, “How to use” (three to four drops in the morning, then moisturizer and SPF), the key ingredients, and three common questions: which size should I choose, can I use it with sunscreen, and is it suitable for all skin types. These are exactly the follow-up questions AI asks itself.
Look at this line in the description: “30 ml ($48, now $42)”. I typed the price into the text. Today it is correct. But the day the sale ends, the price field changes and this text doesn’t. Now the page shows two different prices, and when AI finds contradictions it often cites nobody, or quotes the wrong price.
The rule is simple: never type prices or stock into the description. The price field is the only source of truth.
“Best for: all skin types” doesn’t help AI answer someone looking for a vitamin C serum for sensitive skin. The FAQ does better: if your skin is sensitive, patch test first and start every other morning. Specific and honest wins.
This is the opposite of the old trick from the introduction. If your product is not right for someone, or it doesn’t have a feature, say so clearly. Don’t write around it.
The demo images are stock photos, and the alt text honestly says “Three blank-label dropper serum bottles.” On a real store, use real photos of your real product, with alt text that includes the product name, like “LUMÉRA Vitamin C Radiance Serum, 30 ml glass dropper bottle.”
Suppose a customer asks AI: “I need a fragrance-free barrier cream for dry, sensitive skin, under $40, that works under sunscreen.” Watch how every part of that question matches a clear fact on this page.
“I need a fragrance-free barrier cream for dry, sensitive skin, under $40, that works under sunscreen.”
That’s GEO for products: AI doesn’t have to guess. And an AI agent shopping for this customer can read the price, see the button and add the cream to the cart. That’s AFO.
One FAQ on the page reads: “How long does a 50 ml jar last? Used twice a day, around six to eight weeks.” Most competitors never write a specific, useful fact like this, and AI loves it. The weight and dimensions are filled in too. They help with shipping and make your product data complete for Google Shopping.
What’s missing? Reviews. This product has zero, and in shopping answers both Google and ChatGPT show ratings. Ask every customer for a specific review. “I have dry, sensitive skin and it didn’t sting” is worth ten “great product” reviews, because it contains the exact words the next customer will type.
This is where most stores quietly fail. On this product, Size, Skin Type and Texture are global attributes. They are created once under Products → Attributes, so the spelling is the same on every product and customers can filter by them. But Key Ingredients, Routine Step and Fragrance were typed as custom attributes, only for this one product.
Custom text can be spelled differently on every product: “Fragrance-free” here, “Fragrance free” there, “No fragrance” on a third product. For you, it’s the same thing. For a machine, these are three different facts.
Clean attributes checklist
Managing hundreds of products? Our guide to WooCommerce product specifications at scale shows how to keep attributes clean across a large catalog.
Google’s own AI search guide says Merchant Center feeds help your products show up in AI answers. Connect your store to Google Merchant Center (it’s free) and make sure the feed has the same names, prices and sizes as your website. Do the same for your social shop and any marketplace. A messy feed is the same problem as a messy page.
Structured data, the code that tells machines “this is a product, this is the price, it’s in stock”, is output by WooCommerce itself, without any SEO plugin. One catch: if your product pages are built with Elementor or another page builder, check that it is still there. On this store, one tiny custom plugin puts it back. Check yours in Google’s Rich Results Test. Google also says there is no special AI schema you need to add.
Last thing: the blog post links to this cream from its dry-skin and sensitive-skin sections. Your blog brings the question, and your product page closes the sale.
A perfect page alone doesn’t get you recommended. Next: how AI actually decides who to recommend, the eight things it looks at, and why what others say about you beats anything you say about yourself.
AI fans a question out into several searches, reads mostly the top of each page it finds, writes one answer and cites a few sources. To be recommended, you must be reachable and findable, answer the exact sub-question near the top, be specific, have independent sources that agree with you, never contradict yourself, and be fresh and trustworthy.
Let’s follow one question from start to finish. A customer asks: “What’s a good barrier cream for sensitive skin?” AI decides it needs fresh information. It fans the question out into several small searches, pulls pages from its search index, reads mostly the top part of each page, writes one answer and shows a few sources.
Three different things can happen to your brand in that answer. A citation: your page is shown as a source, usually with a link. A mention: your brand is named, often without any link. A recommendation: AI tells the customer to pick you. The recommendation is the one that makes money.
Is a mention without a link worth anything? Yes. Most AI mentions don’t have a link at all. Mentions build the recommendation, and links bring the clicks you can measure, so you need both.
Every platform also has its own sources. Ahrefs looked at the fifty most-cited websites on Google’s AI Overviews, ChatGPT and Perplexity, and only seven appeared on all three. Don’t optimize for just one platform. Do the fundamentals, and they help everywhere.
Simon from Question Marketing says something I agree with: AI largely decides which brands to recommend from what it already learned, and then looks for sources that support that choice. That’s why what the internet says about you matters so much.
These come from my own experience and everything I studied.
The bot is allowed in, the page loads fast, and the text is in the HTML, not hidden behind JavaScript.
Your page shows up in the search that AI runs. That’s normal SEO, and it’s the foundation.
Not the topic in general, but the specific follow-up: the price, the skin type, “does it work under sunscreen”.
It sits near the top, clear and in simple sentences, like the answers we wrote in Parts 2 and 4.
Numbers, names, sizes and prices.
This one has a dark side. Ahrefs invented a fake brand and planted contradicting stories about it on a blog, on Reddit and on Medium. Some AI platforms repeated those made-up details as facts in more than a third of their answers. When AI has to choose between a vague truth and a specific lie, it often picks the specific lie. ChatGPT did best, and most of the time it cited the brand’s official FAQ.
Publish your own specific, official facts before someone else fills that gap for you.
Several independent sources describe you the same way. If one source says you are great, that’s a claim. If ten sources say it, for AI, that’s a fact.
Simon had a client that described itself as a “Meta business partner” on one page and “part of the Meta business platform” on another. For us that’s almost the same. AI saw two different claims and cited neither.
Same name, same price, same address and same description, everywhere.
Recent updates, a real author and real reviews.
Ahrefs found that 76 percent of ChatGPT’s most-cited pages were updated in the last thirty days.
Notice something? Criteria one to five and seven are on your own website. You control them, and after Parts 2 and 4 our pages pass them. But number six, other people agreeing with you, happens outside your website, and that’s exactly where most stores lose.
Next: on-site versus off-site, how to get people talking about you without spamming Reddit, and whether keyword-stuffed “hyper-SEO’d” content still works in 2026.
On-site work is your blog, product, FAQ, shipping and returns pages. Off-site work is everywhere else AI reads about you: YouTube, Reddit, reviews and “best of” lists. Earn mentions honestly under your real name, track AI’s hidden influence with a “How did you hear about us?” question, and replace keyword-stuffed content with something only you can publish.
On-site is everything on your own website: blog posts, product pages, and your FAQ, shipping and returns pages. Don’t ignore those boring pages. When someone asks AI “Does this store ship to Canada?” or “Can I return an opened product?”, that is where the answer comes from.
Off-site is everywhere else AI reads about you, and it reads a lot. On Google’s AI Overviews, YouTube is the most-cited website. Ahrefs found that YouTube mentions had the strongest link to ChatGPT visibility of any factor it studied, although other experts found that ChatGPT almost never links to YouTube videos in their tests. So YouTube is a must for Google’s AI and one of several channels for ChatGPT. Reddit is cited everywhere, and a big share of cited pages are “best of” lists and comparisons.
A customer sees your brand in ChatGPT and doesn’t click. Later, they type your name into Google, or your web address straight into the browser. In your analytics, that shows up as branded search or direct traffic. AI did the work and gets zero credit.
Add one question at checkout or in your order email: “How did you hear about us?”, with “ChatGPT or another AI” as an option. For Google’s AI, Search Console now has a free Generative AI performance report. You don’t need a paid tool for this.
On Reddit, use your real account, say who you are and where you work, and then give a genuinely useful answer. Ethan Smith says even five good comments in the right threads can be great. People who create a hundred fake accounts get banned.
Some people say Reddit is only cited because it signed paid deals with Google and OpenAI. Those deals are real, but for you the result is the same: Reddit is cited a lot, so being helpful in the right threads still pays.
Fake mentions don’t work. Under one of the videos I studied, I found eleven comments praising the same AI tool in almost the same words. Platforms catch that, and even Google’s own guide says fake brand mentions are not as helpful as people think.
On YouTube, make simple, helpful videos that answer real questions. Put the keyword in the title, add chapters, and say the keyword out loud, because AI reads the transcript. It doesn’t watch the video. Say your brand name out loud too.
For “best of” lists, find the lists AI already cites for your category and give the writer something new: a product sample, your data or a real story. Never pay for placement that reads like an ad. If it reads like you paid for it, assume a machine can tell.
Keyword-stuffed pages, three-thousand-word articles that say nothing, AI rewrites of whatever already ranks: in 2026, none of it helps.
Ethan Smith calls this “typical” content. SEO tools tell everyone to cover the same points, so everyone rewrites each other’s article, and Google and AI keep getting better at ignoring copies. His team checked thousands of Google results and ChatGPT citations. Only about 10 to 12 percent of what ranks or gets cited is AI-generated, even though the internet now has more AI-written content than human-written content.
That doesn’t mean AI writing is banned. One viewer told me he has hundreds of AI-written pages ranking very high, and I believe him, because he also said they are well written and actually help the reader. Google doesn’t penalize content because AI helped. It penalizes thin, spammy content. AI-assisted and human-checked works. Raw AI output published as it is doesn’t.
The slop test (from Rob Hoffman)
Take your page and delete everything that is yours: your data, your photos, your experience, your opinion. If what’s left is the same as every other page on the topic, it’s slop. The fix is to add something true that nobody else has published: your test results, the questions your customers really ask, a photo of the actual product. AI can’t invent these. Only you have them.
Startups don’t have the authority to win Google’s big keywords, but they can own a very specific question that nobody else answers well. As Ethan put it, you could be the only citation for that question and win that whole tail.
For stores, shopping is moving inside AI. Both Google and ChatGPT now show product cards with prices, ratings and links to sellers, and Google says its AI answers can include product listings from Merchant Center. Clean attributes, a clean feed, real reviews and structured data get you into those cards. For most stores this is still early, and a lot of shopping still goes through Google Shopping and Amazon. But early is exactly when you want to be there.
And for everyone, AI has become the last check before people buy. Someone sees your ad or your post, then asks ChatGPT, “Is this brand any good?” If AI knows you and speaks well of you, every other channel you pay for works better.
One thing can make all of this worthless: AI bots that cannot get into your website. Next is the default setting I warned you about at the start, and the plugin question: Yoast, Rank Math, or nothing.
Check your Cloudflare AI crawler settings and your robots.txt, make sure your text loads without JavaScript, keep pages fast with clean headings, run Chrome Lighthouse’s Agentic Browsing check and verify your structured data. An llms.txt file takes ten minutes but Google doesn’t need it. And you don’t need an SEO plugin or an AI plugin.
If your website runs through Cloudflare, go into its bot settings and check the AI crawler options yourself.
On some setups, AI crawlers are blocked, or told to stay away, by default. Ahrefs and Simon both warn about this, and Simon calls Cloudflare “probably your biggest danger.” You built perfect pages, but the audience isn’t allowed in. Make this choice yourself. Don’t let a default make it for you.
Open yourdomain.com/robots.txt and look for GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended. If any has “Disallow” next to it, that bot is blocked.
Ahrefs found that around 6 percent of 140 million websites block OpenAI’s crawler, and many of those blocks come from old templates and defaults nobody remembers adding. Checking takes five minutes.
AI companies use different bots for training and for live search, so you can block the training bot and still allow the search bot. Remember that robots.txt is a request, not a lock. For a hard block, use your firewall or Cloudflare. And don’t block everything: if AI cannot read you, it recommends your competitor instead.
Switch off JavaScript in your browser and reload your product page. If the text disappears, you disappear too.
According to Ahrefs, ChatGPT’s crawler doesn’t run JavaScript, so content that only appears after scripts load is an empty page to it.
Slow pages can be dropped before AI even reads them. Use your headings in order: one H1, then H2s, then H3s.
For a deeper look, see our guide to WooCommerce performance beyond page speed.
Chrome’s Lighthouse test now has an Agentic Browsing section that checks whether AI agents can use your page: real buttons, proper labels and a stable layout.
It’s free and built into Chrome. Run it on your product page and fix what it shows. This is exactly what Google’s agent-friendly guide talks about.
Schema tells machines “this is a product, this is the price, this is the brand.” It helps, but it can’t replace content.
“The schema is going to describe the content. It can’t be a substitute for it.”
Schema on a thin page does nothing.
An llms.txt is a plain text file at yourdomain.com/llms.txt that gives AI a short summary of your site and links to your key pages. Google Search doesn’t need it, but it takes ten minutes and won’t hurt.
Google’s own AI search guide says Google Search doesn’t need llms.txt, markdown copies or any special AI files. But Chrome’s Lighthouse now checks for an llms.txt in its Agentic Browsing section, because other AI agents may use it. It’s not magic, so if someone wants to charge you for it and calls it AEO, walk away. ChatGPT or Claude can write it for you for free. Do it, then forget it. What AI definitely reads today is your normal HTML page, and that’s where your effort should go.
No. The demo store’s plugins page has no Yoast, no Rank Math and no AI SEO tool.
WordPress does the work. It uses your post title as the page title, adds canonical tags, creates an XML sitemap and serves a robots.txt, and WooCommerce outputs the product structured data. The only extra on this store is one small custom plugin. It keeps WooCommerce’s product data on the Elementor product pages and adds a meta description and social preview tags. It does one small job; it is not an SEO plugin.
What do you lose without an SEO plugin? Mainly an easy meta description box. Google rewrites a lot of meta descriptions anyway, and AI answers don’t depend on them, so it’s nice to have, not a must.
No paid tools either. Search Console, Bing Webmaster Tools, Chrome’s Lighthouse and ChatGPT are my whole toolkit for this, and all of them are free. Even Google’s guide warns about tools that claim they can guarantee your ranking.
My honest opinion on the two big plugins. Yoast: I don’t recommend it; in the AI era I haven’t had a good experience with it. Rank Math: if you really want a plugin, it’s comparatively fine. The 94 score at the top of this guide is Rank Math’s score on my own blog. It’s a useful checklist, but a page ranks because of what is on the page, not because of the plugin.
If you do use an SEO plugin, check three things
You have the whole system now. Next, everything in one place: thirty tips grouped by the five A’s. Watch the last one, because almost nobody does it.
These thirty tips are the whole guide in one place, grouped by Access, Answer, Attributes, Authority, Amplify and Measure. If you do only one thing, do tip 30: ask the AI tools your top customer questions every month, write down who gets named, and fix what they say is missing.
Open your robots.txt today, and make sure GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot are not blocked by accident.
If you use Cloudflare, check the AI crawler settings yourself.
Switch off JavaScript and reload your key pages. If the text disappears, fix it.
Make your pages fast: WebP images, compressed, and good Core Web Vitals.
Add your site to Google Search Console and Bing Webmaster Tools. Both now show how you appear in their AI features, for free.
Run Chrome’s Lighthouse Agentic Browsing check on your product page: real buttons, clear labels, a stable layout.
AI sometimes invents web addresses. Ahrefs found AI assistants send people to broken pages almost three times more often than Google does, so check your 404s and redirect the ones getting visits.
One main keyword per page, with the same words in the title, the slug and the H1.
Answer the question in the first sixty words.
Write your headings as the questions customers actually ask, and start every section with its answer.
Put key takeaways at the top, and a comparison table wherever you compare options.
Add an FAQ to almost every page: an H2 for the section, an H3 for each question, two or three sentences per answer.
Find follow-up questions from real customers, not from tool lists, and answer them on the same page, not on a new page for each one.
Make every product’s short description an answer: what it is, who it’s for, the key ingredients or specs, and the size.
Use global attributes with one clean value per field, and fill in the brand, GTIN, weight and dimensions.
Never type prices or stock levels into descriptions.
Never write around features you don’t have. Say clearly what your product does, and what it doesn’t.
The two-second rule: a customer should see what it is, the price and the add-to-cart button right away.
Keep your name, prices and descriptions the same everywhere: your website, your Merchant Center feed, your social shop and any marketplace.
Ask for specific reviews: the customer’s skin type, their problem and their result.
Answer Reddit questions honestly, with your real name and your company. A few good answers beat a hundred fake accounts.
Make simple YouTube videos for the questions people ask: keyword in the title, chapters in the description, and say the keyword out loud.
Ask “How did you hear about us?” at checkout, with “ChatGPT or another AI” as an option, and check Search Console’s Generative AI report every month.
Tip 30 · the one almost nobody does
Once a month, ask ChatGPT, Gemini, Claude, Perplexity and Google’s AI Mode your top ten customer questions. Ask each one more than once, because the answers change, and write down who gets named. Then ask the AI directly: “Why didn’t you mention my store?” Many times it will tell you what is missing. Fix that gap, and check again next month.
This last tip turns everything in this guide from a one-time job into a system that keeps getting better.
If you only have one hour today: check your robots.txt and Cloudflare settings (10 minutes), rewrite your best-seller’s short description and remove typed prices (20 minutes), add “How did you hear about us?” to checkout with ChatGPT as an option (10 minutes), and ask four AI tools your top five customer questions (20 minutes).
That’s the whole system. Access: let the bots in. Answer: put the answer first. Attributes: clean, specific product facts, and no writing around features you don’t have. Authority: show the real person and the real experience behind the page. Amplify: get other people talking about you.
One page, built the right way, for Google and for AI. No SEO plugin, no AI plugin, no paid tool, and no tricks that stop working in six months.
If you are a new business, this is your moment. The big old websites won the old game because of history. In AI search, the store that gives the clearest, most specific and most honest answer can win, even if it launched last month. So think like your customer, not like a marketer.
Want every point in this guide as a list you can tick off? Open the full checklist.
Tell me in the YouTube comments what your store or business is, and the one question you want AI to recommend you for. I read all the comments, and I will answer the best questions in a follow-up video.
Introduction
Part 1
Part 2
Part 3
Part 4
Part 5
Part 6
Part 7
Part 8
Wrap-up
SEO is visibility in search engine results. AEO (answer engine optimization) makes your content the direct answer in tools like Google’s AI Overviews, ChatGPT or voice assistants. GEO (generative engine optimization) gets your brand named inside AI-written answers, and AIO is a loose label for “AI optimization”. Only SEO is an agreed standard, so ask anyone selling the others exactly what work they will do.
It is mostly the same work. AI answers are built on search results, and Google says its AI features don’t need a separate set of technical optimizations. What AI adds is extra weight on brand mentions across the web, clear answers near the top of the page, and consistent facts everywhere.
A plugin score measures the plugin’s own checklist, not Google’s view of the page. If the page lacks value, such as a direct answer, real facts and real experience, or the search has stronger competitors, it won’t rank. On my own site, a 94-score post got one impression, while a 92-score post with real value sits on page one and is cited in AI Overviews.
Access lets search and AI bots read your pages. Answer puts the answer in the first lines. Attributes gives products clean, specific facts. Authority shows the real person and proof behind the page, and Amplify gets other people talking about you. It’s my practical framework, not an official standard, and one set of work covers Google and AI.
Most AI answers are a summary of a search. The assistant runs several searches, reads the results and cites some of them, so if your page isn’t in those results it can’t be in the answer. A Google ranking doesn’t guarantee a citation, but it is the foundation.
ChatGPT runs its own fan-out searches, reads different sources and weighs what others say about each brand. Your competitor may have more independent mentions, more specific facts or more consistent information. Look at the sources the answer cites, then close those gaps.
Yes, especially for specific questions. AI builds answers from many sources, and in Ahrefs’ study of 75,000 brands, brand mentions were more strongly linked to AI Overview visibility than backlinks. A small site wins by answering a narrow question better than anyone else, not through formatting tricks alone.
AI answers are probabilities, not rankings. They change with the wording, location, conversation context and the sources found at that moment. Ask the same questions several times on each platform and track how often you’re named, instead of trusting one screenshot.
Plan for two to three months, with a check every month. A specific question can be cited within thirty to sixty days, but answers vary from run to run, so judge from repeated checks. For Google, watch impressions in Search Console, because they move before clicks.
No. Nobody controls how these systems answer, and Google itself warns about tools and services that promise guaranteed rankings. Ask an agency for defined work, transparent measurement and realistic expectations instead.
More content is more words. Value gives the reader something they didn’t have: a direct answer, original testing, real numbers, a clear comparison or first-hand experience. Use the slop test: delete everything that’s yours, and if the rest matches every other page, add value before you publish more.
Yes, if they are accurate and genuinely help the reader. Google penalizes thin, spammy content, not AI assistance. Add what only you have, such as your data, photos and your customers’ real questions, before you publish.
Use one main keyword in the title, slug and H1, turn real customer questions into H2s, and answer the main question in the first sixty words. Add an author, dates, key takeaways, a comparison table, useful images and a short FAQ. Let the length come from answering more questions, not from a word-count target.
Collect them from support emails, WhatsApp chats, reviews, Reddit threads and Google’s “People also ask” box. You can ask ChatGPT what a customer would ask about your topic, then check the suggestions against real customers. Answer the follow-ups on the same page.
Give each heading one distinct question and merge questions that need the same answer. Write headings in your customers’ words, and don’t force the main keyword into every one.
Readers decide in the first few lines whether to stay, and AI reads the top of a page far more than the bottom. A widely quoted figure says 44 percent of ChatGPT’s citations come from the first third of a page. Answer in the first sixty words, then add the details.
Add a comparison table whenever you compare options, key takeaways at the top of longer pages, and a short FAQ for follow-up questions the main sections don’t cover. Add each one because it helps the reader, not because a template says so.
They let customers and AI match your page to a precise question, like “under $40” or “for sensitive skin”. Vague claims can’t be matched. Keep the facts accurate and official, because AI often repeats whichever specific details it finds.
They are evidence beyond your own claims. When several independent sources describe you the same way, AI treats it as a fact rather than marketing. Earn them honestly through helpful answers, real reviews and genuine coverage; no number of mentions guarantees a recommendation.
Compare your website, Google Business Profile, Merchant Center feed, social shop, marketplaces and directory listings. Fix any differences in your name, prices, address, specifications and policies, starting with the sources customers use most. When AI finds contradictions, it often cites nobody.
They show who stands behind the information, whether it is current and what customers experienced. Link the author to a real profile, update pages every one to three months with real changes rather than a new date, and ask customers for specific reviews.
In two or three plain sentences, say what the product is, who it’s for, the key ingredients or specs, when to use it and the sizes. Replace vague claims like “premium quality” with facts customers can check.
Create shared properties such as size, skin type, texture and fragrance as global attributes under Products → Attributes, so the name, spelling and units are identical on every product and customers can filter by them. Keep one fact per field, use your customers’ words, and keep marketing phrases out of the values.
Typed prices go stale when a sale ends or stock changes, and then the page shows two different prices. AI may quote the wrong one or cite nobody. Let the price and stock fields be the only source of truth.
Say who the product suits, who should choose something else and which features it doesn’t have. Clear limits help buyers choose correctly, cut returns and bad reviews, and let AI match you to the right questions.
Keep product pages public and fast, with a clear name, a short description that answers, real prices per variation, clean global attributes, real photos and specific reviews. Connect a Merchant Center feed that matches the site. No single plugin or submission guarantees inclusion.
It’s usually a good start, because WooCommerce outputs product, price and stock data without any SEO plugin. Page builders like Elementor and extra plugins can remove or duplicate it, so test real product URLs, including variations, in Google’s Rich Results Test.
Yes, if you sell physical products. Google says Merchant Center feeds help products appear in its AI answers, and the feed and the page should show the same names, prices and sizes. A messy feed is the same problem as a messy page.
Never type prices or stock into descriptions, keep the page, structured data and feeds in sync, and clear stale caches. You can’t force every AI tool to refresh instantly, but consistent data removes the conflicts that cause wrong answers.
No. On-site AI search helps visitors find products. External assistants discover you through public pages, feeds and what other websites say, so make sure every product can also be reached through normal links.
Open yourdomain.com/robots.txt and look for “Disallow” next to GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended. If you use Cloudflare or a firewall, check its AI crawler and bot settings. Then switch off JavaScript and reload your key pages to confirm the main text is in the HTML.
Yes, for providers that use separate bots. OpenAI uses OAI-SearchBot for search and GPTBot for training, so you can allow one and disallow the other in robots.txt. Check each provider’s documentation, and remember robots.txt is a request; use a firewall for a hard block.
Yes. Cloudflare, security plugins and bot challenges can block or challenge AI crawlers, sometimes by default. Review blocked requests and bot settings, and allow the verified crawlers you want instead of switching protection off.
No. Google says its search doesn’t need llms.txt, and support among other AI tools varies. It takes ten minutes and won’t hurt, but your normal HTML pages are what AI reads today.
No. WordPress and WooCommerce already output most of what search engines and AI need. Use a plugin only to solve a specific problem you have found, and never run two SEO plugins at the same time.
WordPress already handles page titles, canonical tags, an XML sitemap and robots.txt, and WooCommerce outputs product structured data. You still need to check indexing, redirects and the schema on page-builder templates, and add meta descriptions if you want control over them. Audit the real output instead of assuming.
In GA4, open the Traffic acquisition report and use the Session source / medium dimension to find referrers such as chatgpt.com, gemini.google.com and perplexity.ai. Many AI-driven visits arrive as direct or branded search instead, so also ask “How did you hear about us?” and check Search Console’s Generative AI report.
Use your real account, follow each community’s rules, say who you are and where you work, and give a genuinely useful answer. A few good comments in the right threads beat a hundred fake accounts, which get banned.
Find the claim and the source it cites. Correct your official pages, publish a brand FAQ with specific facts, ask other publishers to fix their errors and use the platform’s feedback option. Recheck every month, because corrections take time to appear.
Check robots.txt and Cloudflare (10 minutes), rewrite your best-seller’s short description and remove typed prices (20 minutes), add “How did you hear about us?” at checkout (10 minutes), and ask four AI tools your top five customer questions (20 minutes). Note today’s impressions and sales so you can compare later.
Written by
Head of Engineering, Rowerr AI & IT Services
Abdul Wahab has worked on websites, mobile apps and their SEO since 2014. At Rowerr, his team builds WooCommerce stores, membership sites, LMS platforms and AI workflows, with more than 1,100 projects completed since 2019. Everything in this guide comes from that client work and his own sites.
Schedule a strategic consultation to align your goals with a solution built for long-term success.
