Answer Engine Optimization (AEO): What It Is and What to Do
You have been told to "do AEO". A vendor sent a deck and a peer quoted a stat. The advice at the top of Google reads the same on every page: write in questions and answers, answer in the first 40 to 60 words, add FAQ schema, publish where the AI looks. All of that is true. None of it is the reason the AI names your competitor and not you.
Answer engine optimization for a business owner is a trust problem before it is a content problem. ChatGPT, Gemini, Perplexity and Google's AI Overviews name the businesses they can verify from sources that are not you. The formatting advice is table stakes. The deciding factor is your third-party record: reviews, listings, mentions, and a name that means one company. I have spent 17 years in search, worked with 300+ businesses, and I now track how the AI engines answer questions about client businesses, and what Google does with my own site. This page is what that tracking taught me, in the order I would do the work if the business were mine.
What answer engine optimization actually is
Answer engine optimization (AEO) is the work of getting your business named, cited or recommended inside the answer an AI system gives, instead of ranking as a link underneath it. The answer engines are ChatGPT, Gemini, Perplexity, Claude, Copilot, and the AI Overviews and AI Mode inside Google Search. The measurement is no longer a ranking position. It is whether the answer mentions you, whether it links to you, and whether it puts you ahead of the competitor the customer would otherwise call.
That definition splits the work into two jobs. Publishers want to be the cited source for a fact. Business owners want to be the recommended business for a need. The pages ranking for this term are written for the first job. If you run a clinic, a contractor business, a law firm or an online store, you are doing the second job, and the second job has a different bottleneck. I cover the comparison itself in how AEO and SEO differ in 2026. This page is about what to do.
| For a business owner | SEO | AEO and GEO |
|---|---|---|
| What you are optimizing | A page, to rank for a query | A business, to be named in an answer |
| Where it shows | The list of links, the map pack | AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity, Copilot |
| What moves it most | Relevance, links, technical health | Facts about you that the engine can verify across sources you do not own |
| How you measure it | Position, clicks | Mentions, citations, who gets named instead of you |
AEO (answer engine optimization) and GEO (generative engine optimization) are used for the same job by most of the market. This page treats them as one.
Why the answer, and not the link, is now the prize
The clicks have moved. Pew Research Center tracked the real browsing of 900 US adults in March 2025. People who saw an AI summary clicked a traditional result on 8% of visits, against 15% when there was no summary. Clicks on a link inside the summary itself happened on 1% of visits. Searches phrased as questions produced a summary 60% of the time. The customer who asks a question is the customer who gets an answer, and the answer names someone.
The other side of the shift is who is asking. Profound analyzed 7.5 million English-language ChatGPT conversations from June 2025 to June 2026 and found the share with commercial intent rose from 13.9% to 19.2% in a year. Home services sat at 30.3% commercial. Healthcare and legal stayed informational-first, which for a clinic or a firm means the first question is about the problem and the second is about who to call.
| What the numbers say | Finding | Source |
|---|---|---|
| Clicks when an AI summary appears | 8% of visits, against 15% without a summary | Pew Research Center, July 2025 |
| Question searches that trigger a summary | 60% of searches starting with who, what, when or why | Pew Research Center, July 2025 |
| Commercial share of ChatGPT conversations | 13.9% to 19.2% in one year; home services 30.3% | Profound, August 2026, 7.5M conversations |
Sources: Pew Research Center, "Google users are less likely to click on links when an AI summary appears in the results" (July 22, 2025); Profound, "Commercial conversations in ChatGPT more than doubled in a year" (August 13, 2026).
How the engines decide who to name
Google describes the mechanism in its own documentation. AI Overviews and AI Mode may use a "query fan-out" technique, issuing multiple related searches across subtopics and data sources before writing the response. A page needs no special optimization to be eligible. It needs to be indexed and eligible for a snippet, and Google says plainly that there are no additional technical requirements. ChatGPT and Perplexity do a version of the same thing: they search, read what comes back, and write.
Now picture the fan-out for "best physio clinic in Leeds for a running injury". The engine runs searches for reviews, for "best physio Leeds" roundups, for the directory and health-register listings, and for the clinic sites themselves. A clinic with a beautiful site and no footprint anywhere else appears in one of those searches. A clinic with 200 detailed reviews, three roundup mentions and one consistent name appears in all of them. The engine names the second clinic. It is doing what a careful person would do with the same evidence.
What my own tracking shows
In July I ran ten category questions about a real estate client in Virginia through Claude, three times each. "Who is the best real estate agent in Hampton Roads?", "Best real estate company for military relocation in Hampton Roads", and eight more in that shape. Thirty answers. The client, a well-built site with the right pages, was named in none of them. The answers named the same three regional brokerages every time, the ones with the deepest record of reviews, listings and local press.
Then there is my own site. I did the formatting work by the book: 115 posts, each with a question-led FAQ and FAQPage schema, answers up front, a named author box, a PageSpeed score of 100. What I did not have was a third-party record. The domain rating sits at 0. The result over the 90 days to September 1 was 1,230 impressions and 17 clicks in Google Search Console. The structure got the pages indexed and read. It did not get them trusted, and trust is what an answer engine spends when it names a business.
One more from the tracker. When I asked ChatGPT about a junk removal client by name, it stopped to ask which of the several companies operating under that name in different states I meant. The engine was not being difficult. Nothing beyond the client's own site told it which one was real. That is an entity problem, and no FAQ block fixes it.
The order of work
Everything above points at one sequence. Do the trust work first, because it is slow and it compounds. Then make the site easy to quote. Then measure the answers, because rankings will not tell you what the engines are saying.
The answer block, which everyone teaches and which still matters
Once the record exists, make each page easy to lift from. This is the part the guides get right, and it is quick. HubSpot's guide tells you to put the core answer in the first 40 to 60 words. CXL looked at 100 Google AI Overview citations and found 55% came from the first 30% of the cited page. Both match what I see on client pages: the engines quote the top, and they quote the sentence that reads as a complete answer on its own.
What not to buy
Some things sell well on this topic and do little. A tracking tool before you have anything to track: ten questions in a spreadsheet, asked monthly, will tell you the same thing for free until you have a record worth measuring. A paid placement in a scheme that promises to get you "into ChatGPT": nobody sells that access, and I have watched businesses pay for it. And llms.txt, the text file vendors keep recommending. Google's own guidance on appearing in AI features lists no such requirement. My own file has redirected to my homepage for months with no visible effect either way.
Spend that money on the record instead. A reviews process that produces detailed reviews every week, a listing in the two roundups your customers actually read, a named expert page. Those show up in the fan-out. A dashboard does not.
Picture your own business
Picture a dental practice with two locations and an implants service it wants to grow. A customer asks ChatGPT for the best dentist for dental implants in Manchester. The fan-out searches hit review platforms, "best implant dentist Manchester" roundups, the regulator's register, and the practice sites. One practice has 400 reviews that mention implants by name, an implantologist with credentials on a page of her own, listings in two local roundups, and an implants page with a price range. The other practice has a prettier website. The first practice gets named. The second one does not appear, and its owner concludes that AI search is not working for dentists.
The fix for the second practice is the sequence above, in that order. Make the name consistent everywhere. Get the reviews flowing with detail. Get onto the two roundups. Put the price range and the named implantologist on the implants page. Then format the page as an answer block. Six to twelve weeks of work, most of it away from the website.
How to measure it without fooling yourself
A single ask tells you nothing, because the same question gets different answers on different days. Ask ten category questions, three times each, monthly, across the engines your customers use. Log every business that was named instead of you. That list is your outreach target list: the roundups, directories and review platforms those businesses share are the fan-out searches you are missing. I walk through the exact procedure in how to check whether AI recommends your business.
Then watch the referrals. Traffic from ChatGPT, Perplexity and Gemini shows up in analytics with its own referrer when the click is clean, and often as direct traffic when it is not. Google counts AI Overviews and AI Mode inside the normal Search Console data, and it added a Generative AI performance report in June 2026. The number to watch is not the click count. It is whether the questions that used to name a competitor now name you. For the wider picture, from how the engines pick sources to what the buyer journey looks like now, the complete guide to AI search visibility goes layer by layer.
Frequently Asked Questions
What is answer engine optimization in plain English?
Answer engine optimization is the work of getting your business named or recommended inside the answer that ChatGPT, Gemini, Perplexity or Google's AI Overviews give a customer. Ranking as a link under that answer is a different job. The engines build those answers by searching for your business across reviews, listings, mentions and your own site, and they name the businesses they can verify. For a business owner, the work is mostly building that verifiable record, then formatting your pages so that the engines can quote them.
Is AEO different from SEO, or is it the same work with a new name?
The foundations are the same: an indexed, fast, honest site with clear pages. Google says there are no additional technical requirements for its AI features. What changes is the unit and the bottleneck. SEO optimizes a page to rank for a query. AEO optimizes a business to be named in an answer, and for most businesses the bottleneck is the third-party record, not the page. If your SEO is in good shape, AEO is mostly reviews, listings, mentions and a consistent name, plus an answer-first format on your key pages.
How long does answer engine optimization take to work?
The formatting changes show up within weeks in how your pages get quoted for questions you already rank for. The trust record takes months, because reviews, listings and mentions accumulate at the speed of your customers and the publications you approach. For category questions like "best physio in Leeds", expect three to six months before the answers start naming you, and measure monthly with the same ten questions so you can see the movement.
If you want to know where your business stands right now, I run the ten questions for you as a $150 AI visibility audit. Four engines, three runs, and the list of who the engines name instead, with the record each one has. Book a free AI visibility call and bring the questions your customers actually ask.
Or start with the free AI visibility checker: nine signals, two minutes, and the first thing to fix.
