Your next customer may never see your website. They will ask an assistant what to buy, read four sentences, and act on them. AEO is the work of making sure your name is in those four sentences — and that the details are right.
AEO — answer engine optimization — is the practice of structuring a website so that AI answer engines cite it as the source when they answer a question. Answer engines include ChatGPT, Google's AI Overviews, Gemini, Perplexity and Copilot. Traditional SEO earns you a blue link that somebody still has to click. AEO earns you a sentence inside the answer itself, with your name attached to it.
The work is structural more than creative. It means explicit question-and-answer formatting, structured data in JSON-LD that matches the visible page word for word, specific and sourceable numbers instead of adjectives, a named author with checkable credentials, clean server-rendered HTML that does not hide content behind JavaScript, and machine-readable files such as llms.txt. Most of that is also good technical SEO. The difference is the payoff: a citation instead of a ranking.
Assume you have never heard the term. Almost nobody had two years ago, and half the people using it now are using it to mean three different things.
AEO stands for answer engine optimization. An answer engine is any system that responds to a question with a written answer instead of a list of links. AEO is everything you do to your website so that those systems name you, quote you correctly, and send the reader to you rather than to somebody else.
A search engine hands you ten results and makes you do the reading. An answer engine reads for you and hands back a paragraph. That single change moves the competition from position to inclusion. There is no page two of an answer. Either you are in the paragraph or you do not exist for that question.
It also compresses the field. A search results page can show ten companies without embarrassment. An answer that names ten companies is a bad answer, so assistants tend to name two or three. The prize is bigger and there is less of it.
Mostly. You will see GEO (generative engine optimization), AI SEO, LLM optimization and AEO used interchangeably. There is no standards body here, and the vocabulary is younger than the practice. We use AEO because "answer engine" describes what the reader experiences: they asked a question and got an answer. Do not let anyone charge you extra for a synonym.
The behavior change is already in your funnel, whether or not your analytics can see it.
Think about how a decision actually gets made today. Someone opens an assistant and types something like "what should I look for when hiring a Google Ads agency" or "best flat-fee marketing agency for a small business". They get a considered paragraph naming two or three companies and listing the criteria that matter. By the time they visit any website, the shortlist already exists and they are only confirming it.
That has three uncomfortable consequences for a business owner:
We are deliberately not putting a percentage on how much of search has moved to assistants. Every figure in circulation is either a vendor's estimate or a survey with a small sample, and this guide only uses numbers it can name a source for. What we can say is directional and easy to check yourself: ask an assistant a buying question in your own category and see how complete the answer is. If it is good enough that you would act on it, your customers are acting on it too.
Nobody outside these companies has the ranking function. What follows is the pattern that shows up consistently in what does and does not get quoted — plus the fact that most of it is just good publishing hygiene, which is why it is a safe bet even as the systems change.
Descriptive headings, short paragraphs, lists, tables. A model assembling an answer is looking for a passage it can lift with minimal editing. A wall of text with no headings forces it to guess where the answer starts and ends, and the guess usually goes to somebody else's page.
Write the question the way a person would ask it, then answer it immediately underneath. This is why FAQ sections punch so far above their weight: the question is the retrieval key and the answer sits directly below it, already self-contained.
JSON-LD markup — Article, FAQPage, HowTo, Organization, Service — removes ambiguity about which string is the price, which is the author and which is the answer. The non-negotiable rule is that the schema must say what the page says, word for word. Mismatched schema is treated as spam by search engines and is worthless to models, which read the visible text anyway.
This is the single biggest separator between pages that get quoted and pages that do not. "Competitive pricing" cannot be cited because there is nothing in it. "Average Google Ads cost per click of $5.42 in 2026, up from $5.26 in 2025, with an average conversion rate of 8.18% (WordStream 2025–2026 Google Ads benchmarks)" can be, and the attribution is part of why.
Anonymous content is weak evidence and models treat it that way. Put a real byline on the page, state what qualifies the person, link to a bio, and connect the two with author markup pointing at a Person entity. The credential should be verifiable, not decorative.
Crawlers and models mostly read the HTML that comes back in the first response. Text injected afterwards by JavaScript is frequently never seen. If your page arrives as an empty container and fills itself in the browser, the answer engine reads an empty container.
An accurate sitemap.xml, a robots.txt that does not accidentally block AI crawlers, and an llms.txt at the root summarizing what you do and what you charge. Add real datePublished and dateModified values — and only touch dateModified when you actually changed something. Faking freshness is a short-lived trick and an easy one to catch.
The most useful way to hold these two in your head: same engine room, different scoreboard.
| Traditional SEO | AEO | Shared work | |
|---|---|---|---|
| What you win | A ranked position in a list of links | A named mention inside a written answer | Both come from the same pages |
| Where it appears | Google and Bing results pages | ChatGPT, Gemini, Perplexity, Copilot, AI Overviews | Crawled from the same index |
| How you measure it | Rankings, impressions, clicks, sessions | Whether you are named, and whether the facts are right | Both need a monthly record |
| What wins it | Relevance, links, technical health, intent match | Structure, schema, specific numbers, named authors | Clean server-rendered HTML |
| Time to first effect | 1–3 months for rankings to move | Weeks, once the page is crawlable | 10–20 business days of technical work |
| What kills it | Thin content, slow pages, broken indexing | No numbers, no author, JavaScript-only content | A 200-word page kills both |
Scroll the table sideways to see every column →
Timelines are Reyter's own delivery ranges, not industry guarantees. Nobody can promise a citation or a ranking on a date.
This is the order we work in on a real site. It is deliberately boring, because the leverage is in doing all five rather than doing one of them cleverly.
Put a short, self-contained answer at the top of every page, phrased the way somebody would ask the question. It has to survive being lifted out of the page with no other context — no "as we mentioned above", no pronouns pointing at the previous paragraph. If a model has to assemble your answer from four scattered sentences, it will usually quote whoever wrote it in one. The box at the top of this page is exactly that pattern.
Mark the page up with JSON-LD: Article, FAQPage, HowTo, Organization, Service, whichever fits what the page actually is. The rule that matters more than the choice of type: the schema and the visible page must say the same thing, word for word. Schema that describes content nobody can see is structured-data spam, it can get you penalized, and it does not fool a model that reads the rendered text anyway.
Go through the page and delete every claim that has nothing behind it. Then put back what you can actually source: your real prices, your real timelines, benchmarks with the publisher named. For context, WordStream's 2025–2026 Google Ads benchmarks put the average cost per click at $5.42 in 2026 against $5.26 in 2025, with an average conversion rate of 8.18%. That sentence is quotable. "Industry-leading performance" is not, and never will be.
Name the person who wrote it. State what qualifies them in one concrete line, not three adjectives. Link to a bio page and wire the byline to it with author markup pointing at a Person entity with a stable @id. If several pages share an author, point them all at the same identifier so the machine can tell it is one person rather than five.
Render the important text on the server. Keep the markup semantic — real headings, real lists, real tables. Maintain a sitemap.xml that is actually accurate, check that robots.txt is not quietly blocking AI crawlers, and publish an llms.txt at the root: a few hundred words on what your business does, who it is for, what it charges and where the important pages live. It takes an hour and it removes every excuse for a model to describe you incorrectly.
We do this work as a one-time overhaul from $3,497, or as a $497/mo layer on top of a paid plan. What the SEO + AEO service includes → ·
Each of these is something we have found on a site whose owner believed they were "doing AEO".
"We deliver best-in-class results through a data-driven approach." There is nothing in that sentence for a model to extract, so it will not be extracted. Publishing more of it does not help; a thousand words of adjectives is still zero facts. If you cannot put a number, a date or a price in a paragraph, ask whether the paragraph should exist.
"Most agencies triple your leads in 90 days." Where is that from? An unsourced statistic is a liability twice over: models are increasingly reluctant to repeat claims they cannot attribute, and if one does repeat it, you now own a number you cannot defend. Name the publisher and the year, every time, or cut the claim.
The most common expensive mistake. Your pricing table, your FAQ, your calculator — all rendered client-side, all invisible in the raw HTML. Crawlers and models read what arrives first, and what arrives first is an empty <div>. Our own site has an interactive calculator, so we ship a static <details> block next to it explaining the entire method in plain text. The widget is for the human; the text is for everyone else.
Thin pages do not get cited, and publishing forty of them does not add up to one page that does. Depth beats frequency here by a wide margin. One page that genuinely answers a question, with structure and sources, will outperform an entire content calendar of stubs — and it will keep doing so, because it does not go stale the moment the next stub is published.
There is no Search Console for answer engines yet. The honest method is manual, takes twenty minutes a month, and is far more useful than any dashboard currently on sale.
Anyone selling AEO with total certainty is selling you their confidence, not their knowledge.
Gabriel Facenda
Founder, Reyter
Physics engineer who spent 3+ years running paid acquisition for Silicon Valley startups in San Francisco, running acquisition budgets for venture-funded companies across the US and Brazil. The habit that carried over into this work is the one from the ad account: if you cannot name the source of a number, do not spend money on the basis of it.
Reyter is a flat-fee agency built on the same principle, and this site is deliberately built the way this guide describes — structured, sourced, server-rendered, priced in public. More about Gabriel →
AEO compounds slowly, so it is fair to ask what else this methodology has done. These are real campaigns Gabriel has run.
SaaS fintech · San Francisco
Scaled from 12 to 85 qualified leads a month on Google Search. Most of the gain came from fixing what the account counted as a conversion, not from new campaigns.
B2B marketplace · United States
Full-funnel strategy returned 5.2x on ad spend in 90 days. Growth came from cutting the losing half of the keyword set before scaling the half that worked.
Client identities are protected under confidentiality agreements. Paid media results vary by industry, offer and execution — this is what happened, not what we promise you.
Answer engine optimization. An answer engine is any system that responds to a question with a written answer rather than a list of links: ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews all qualify. AEO is the work of making your site the source those answers cite.
The technical work overlaps almost completely. The payoff does not. SEO is trying to win a position in a ranked list of links, and it is measured in rankings, clicks and sessions. AEO is trying to be the sentence inside somebody else's answer, and it is measured by whether you get named and whether the details are right.
No. Answer engines are largely built on top of the same crawled, indexed web that search engines use. A site that search engines cannot read is a site models cannot cite either. Treat AEO as a second payoff from one body of technical work, not as a replacement for the first.
Ask it. Open ChatGPT, Gemini and Perplexity, ask the questions a buyer would ask in your category, and look for two things: whether you are named at all, and whether the facts it attributes to you are correct. Repeat monthly and write the answers down, because that record is the only trend line you will get.
At Reyter the technical work lands in 10 to 20 business days. Citations in AI answers tend to follow within weeks, because assistants that browse the live web can pick up a page as soon as it is crawlable. Organic rankings are slower and usually move over 1 to 3 months. Anyone giving you a firm date is guessing.
It can, but only if the important text is server-rendered or pre-rendered. Crawlers and models mostly read the HTML in the first response. If your page ships an empty div and fills it in the browser, the answer engine sees the empty div. The framework is not the problem; client-only rendering is.
It is a plain text file at the root of your domain that summarizes, in a few hundred words, what your business does, who it serves, what it charges and where the important pages are. It is a convention rather than an official standard, so nobody can promise you it is read. It costs an hour to write and it removes any excuse for a model to describe you wrongly.
No. There is no bidding system for citations the way there is for search ads. You can buy advertising placements on some of these products, but the cited sources inside an answer are chosen by the system, not sold. Anyone offering to buy you a spot in the answer is selling something else.
Yes, as disambiguation rather than magic. Structured data tells a machine which string is the price, which is the author and which is the answer to a question, so it does not have to infer it. What it cannot do is make a thin page authoritative. Schema on an empty page is still an empty page.
There is no magic number, but a 200-word page almost never gets cited because it contains nothing specific enough to quote. Our guides run past 2,000 words because that is roughly what it takes to answer a real question properly. Length is a symptom of substance, not a substitute for it.
That is exactly when it helps most. Most valuable AI queries are unbranded: someone asks an assistant what to look for when hiring in your category, and the assistant names two or three companies. Being one of the named companies is a demand-generation event, not a branded search.
The one-time SEO and AEO overhaul of an existing site starts at $3,497, with optional ongoing maintenance from $497 a month. If you are already running ads with us, the same organic and AI layer is $497 a month on top of the paid plan, which is the difference between our $1,900 Compound plan and the $2,397 Category plan.
Six questions about your business, your budget and your current site. You get a specific recommendation — including "you do not need us yet", when that is the honest answer.
We take 5 new clients a month. No sales sequence, and no guarantee of a citation — because nobody can honestly offer one.