AEO in two sentences.
Answer engine optimisation is the practice of making your company easy for AI search tools to find, understand and quote, so that it appears in the answers they write for buyers. It rests on four things you control: the questions you choose to answer, whether engines can read your site, how clearly your pages answer, and what other sites say about you.
An answer engine is any search tool that replies with a written answer instead of a list of links. Your aim is to be one of the companies it names, described accurately, with your own page among the sources.
| Question | Short answer |
|---|---|
| What is it? | Work that gets your company named, described correctly and cited in AI-written answers. |
| Where does it apply? | Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Microsoft Copilot and similar tools. |
| What is the unit of work? | A buyer question, not a keyword. |
| What does it build on? | Ordinary SEO: pages that can be crawled, indexed and understood. |
| What is the result? | How often you are named across a fixed set of questions, and the enquiries that follow. |
This page is the starting point for our AI search guides: each chapter gives the short version and points to a longer one. It is also the method behind our answer engine optimisation (AEO) services. For outside help, see what an AEO agency does week by week.
Which answer engines matter.
For most B2B companies, four families of answer engine matter: Google’s AI features, ChatGPT search, Perplexity and Microsoft Copilot. They all read the public web, and each company documents how its engine finds pages.
| Engine | How it finds your pages | What its maker says |
|---|---|---|
| Google AI Overviews and AI Mode | Google’s normal search index | “To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet.” |
| ChatGPT search | The OAI-SearchBot crawler | “OAI-SearchBot is used to surface websites in search results in ChatGPT’s search features.” |
| Perplexity | The PerplexityBot crawler | “PerplexityBot is designed to surface and link websites in search results on Perplexity.” |
| Microsoft Copilot and Bing | Bing’s index | Bing Webmaster Tools reports citations across “Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations.” |
There is no separate “AI web”: every engine works from pages a crawler fetched and indexed. But the engines do not share one index, so being visible in one does not make you visible in the others.
Google’s AI Mode behaves differently enough from AI Overviews to need its own attention. Our guide to Google AI Mode for B2B sites covers what it is and how to see it in Search Console. We also cover how Perplexity finds and cites sources.
Is AEO different from SEO?
AEO is SEO with a different finish line. The foundations are the same, but the target changes from ranking a page to being named and cited inside an answer.
Google is direct about this. Its guide to generative AI features has a note headed “What about ‘AEO’ and ‘GEO’?” which says that “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” Its AI features documentation adds: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” We go through Google’s AI Overviews eligibility rules in a separate guide.
That guards against secret techniques, but it is not the whole picture:
- The goal is different. A page can rank well and still not be the passage an engine quotes, or the company it names.
- The evidence is wider. Engines draw on what other sites say about you, not only on your own pages.
- The measurement is different. There is no stable “position one” in an answer. You measure how often you appear.
You will also see the term GEO, for generative engine optimisation. It comes from a 2023 research paper, “GEO: Generative Engine Optimization”, and in everyday use it overlaps heavily with AEO. For how the three labels relate, read our comparison of AEO, GEO and SEO as an investment decision. In this guide we use AEO for the whole job.
What you can safely ignore
Google’s guide lists things you do not need to do for its AI features: creating llms.txt files or other special markup, breaking content into tiny “chunks”, rewriting pages in a special style for AI, seeking inauthentic mentions, and over-focusing on structured data. If a proposal leans on those items, compare it with our guide to what AI search work should include and cost. We also explain what a credible AI SEO service includes.
Why B2B firms should care.
B2B buyers increasingly ask an AI tool before they visit any supplier’s website, and when an answer appears they click less. A company missing from the answer can lose the enquiry without ever seeing a visit.
G2 surveyed more than 1,000 B2B software buyers in August 2025 and reported that “half of the buyers said they now start the buying journey in an AI chatbot instead of Google Search.” AI chat had also become the top source for building a software shortlist, ahead of review and vendor websites. That is software buyers, not every B2B sector.
Pew Research Center tracked the browsing of 900 US adults in March 2025. When a Google results page showed an AI summary, people clicked a traditional result in 8% of visits. Without a summary it was 15%. They clicked a link inside the summary itself in 1% of visits.
Keep the scale in proportion. Conductor’s benchmark of 1,215 enterprise domains found that AI referral traffic was 1.08% of all website traffic across ten industries between May and September 2025, with 87.4% of it coming from ChatGPT. AI answers shape which names a buyer has in mind more often than they send a visit. We chart the click data in more depth in our analysis of zero-click searches for B2B.
So the commercial case for AEO is not “more traffic”. It is being on the shortlist a buyer has formed before they visit anyone.
How engines pick and phrase answers.
An answer engine turns a question into several searches, reads pages from its index, writes an answer from the passages it found, and cites a few of them. You can influence every stage except the writing. Our LLM SEO guide covers the two routes into an answer, training data and live retrieval.
One question becomes several searches
Google says both AI Overviews and AI Mode may use a “query fan-out” technique, “issuing multiple related searches across subtopics and data sources” to build a response. A company with a page for each sub-question has more chances to be retrieved.
Engines read more than they cite
OpenAI’s developer documentation for its web search tool separates the pages a model consulted from the ones it credits. It says inline citations “show only the most relevant references”, while the sources list is “the complete list of URLs the model consulted”, and “the number of sources is often greater than the number of citations.” A page can shape an answer without being cited.
Answers vary from run to run
Google notes that AI Mode and AI Overviews “may use different models and techniques, so the set of responses and links they show will vary.” Research by SparkToro and Gumshoe.ai, reported by Search Engine Journal in January 2026, ran 2,961 prompts across ChatGPT, Claude and Google’s AI, repeating each of 12 brand-recommendation questions 60 to 100 times per platform. It found less than a one-in-100 chance that ChatGPT or Google’s AI would return the same list of brands twice. How often a brand appeared across many runs was steadier than its position in any one list. The research was not peer-reviewed.
For a closer look at one engine, see how ChatGPT decides which companies to recommend. For evidence on what gets cited, our study of the sources in 36 AI answers found that company and vendor websites made up about 91% of citations to 12 B2B buying questions.
Step 1: collect buyer questions.
Start AEO with a fixed list of 20 to 30 questions your buyers really ask, not with keywords. That list, a question panel, is both your plan and your measuring stick.
- 01 · ASKBuild a panel of real buyer questions
- 02 · ACCESSMake sure engines can fetch and read your pages
- 03 · ANSWERSPublish a clear, provable answer to each question
- 04 · MENTIONSEarn a place in the sources engines already use
- 05 · MEASURERe-run the panel monthly and track enquiries
Work in this order; each step depends on the one before.
Collect the questions from people, not tools: first sales calls, enquiry forms, lost-deal notes and the support inbox. Then sort them into five types, because each type needs a different page.
| Type | How buyers phrase it | Page that should answer |
|---|---|---|
| Category | “Who are the best providers of X for Y?” | Service page written for that buyer; third-party lists |
| Comparison | “X or Y for a company like ours?” | An honest comparison page |
| Problem | “How do we fix or decide Z?” | A guide with a direct answer first |
| Proof | “Is this company any good at X?” | Case studies, reviews, credentials |
| Facts | “What does this company do, and where?” | About and company-facts pages |
A finance director at a 40-site care group might ask:
“Which payroll providers specialise in UK care homes with shift-based staff?” (category)
“Outsourced payroll or in-house software for 1,200 care workers?” (comparison)
“How do we handle sleep-in shift pay correctly?” (problem)
“Is [provider] reliable for multi-site care payroll?” (proof)
Invented questions for an invented buyer, written to show the pattern. They are not real queries or answers.
Now run every question through the engines, more than once, and record who is named and which sources are cited. Our guide on how to run a manual AI answer test gives the full procedure, and the plays on building a question panel and repeating runs are short versions. To see a sample first, our free AI visibility report runs a set of your buyers’ questions for you.
Step 2: let the engines read you.
Before you write anything new, confirm that each engine’s crawler can fetch your pages and that the important text is in the HTML. A page that cannot be read cannot be quoted.
Google’s list of best practices for AI features begins with “ensuring that crawling is allowed in robots.txt, and by any CDN or hosting infrastructure” and includes “making sure that important content is available in textual form.” OpenAI states the consequence of a block plainly: “Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers”, though they can still appear as navigational links.
Check four things on your ten most important pages:
- ✓robots.txt does not block OAI-SearchBot, PerplexityBot or Googlebot.
- ✓The firewall or CDN does not challenge those crawlers before they reach the page.
- ✓The page source contains the main text, without needing JavaScript to load it.
- ✓The page is indexed in Google and in Bing.
The detail, including whether you need an llms.txt file, is in our guide to AI crawlers, robots.txt and schema. To confirm the crawlers arrive, see how to check your logs for AI bots.
Step 3: write the answer on the page.
For every question on your panel, one page on your site should answer it directly in its first two sentences, then prove the answer. Engines lift passages, so each must make sense alone.
Google’s advice for AI features is to publish “unique, expert-led content that provides value beyond common knowledge”. For a B2B firm that means the specifics only you can state: who you serve, what the work includes, how long it takes and how you approach pricing.
A page that is easy to quote has these features:
- ✓The question is a heading, worded the way the buyer asks it.
- ✓The answer comes first, in one or two plain sentences that name the subject.
- ✓Facts are specific and dated: numbers, places, sectors, time frames.
- ✓Comparisons sit in a table, with honest trade-offs.
- ✓Company facts match everywhere: the same name, services and locations on every page and profile.
We set out the evidence and the layouts in the page patterns AI engines quote. The short play on answer-first sections is a good brief for a writer.
Accuracy is part of this step, because engines repeat out-of-date facts from old pages and profiles. If an engine already says something wrong about you, follow our guide to correcting wrong AI answers at the source before you publish more.
Step 4: earn mentions elsewhere.
Engines lean on what the rest of the web says about a company, so a strong website alone is rarely enough for category questions. The task is to appear, accurately, in the sources an engine already cites for your panel.
Ahrefs studied 75,000 brands in May 2025 and compared several factors with how often each brand was mentioned in Google’s AI Overviews. Mentions on other websites had the strongest relationship, well ahead of backlink counts.
A second Ahrefs study, in December 2025, looked at 750 early-stage prompts in ChatGPT and found that “best X” blog lists were 43.8% of all the page types cited. The lists, directories and review sites already cited for your category questions are the first places to be present.
Google’s guide warns that “seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem”, because its AI features depend on the same systems that block spam. Earn mentions the slow way: accurate profiles, real customer reviews, contributions to trade publications and partner listings.
To start, list the sources cited for your panel in step 1 and mark those that name competitors but not you. Our guide to the off-site sources AI engines rely on ranks them by effort and value, and why brands go missing from AI shortlists gives an order of work.
Step 5: measure it.
Measure AEO with four numbers: how often you are named across your panel, how often your pages are cited, how many visits arrive from AI tools, and how many of those become enquiries. No single tool gives you all four.
| Measure | Source | What it tells you |
|---|---|---|
| Share of answers that name you | Your question panel, re-run monthly | Whether buyers hear your name. Count appearances across runs, not position. |
| Impressions in Google’s AI features | Search Console’s Search Generative AI performance report | “How often URLs from your site appeared in generative AI features in Search and Discover.” Rolled out to all websites as of 31 August 2026. |
| Citations in Copilot and Bing | AI Performance in Bing Webmaster Tools | Total citations, cited pages and a sample of the queries used. Bing says it does not indicate ranking or authority. |
| Visits from AI tools | The AI Assistant channel in Google Analytics 4 | Arrivals “from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok”. Google’s AI Overviews and AI Mode are counted under Organic Search instead. |
| Enquiries and pipeline | Your forms and CRM | Whether being named leads to a conversation worth having. |
The first row matters most and is the one you must build yourself. Report the percentage of runs in which you appear, and keep the panel fixed so that months are comparable. We explain how an AI visibility score is calculated in a separate guide.
For the one-page view, see reporting AI visibility to a board. For the tracking set-up, see tracking AI referrals through to pipeline. To judge whether the effort pays, use the AI visibility ROI model. If you are weighing software for the monitoring, we have compared the main AI visibility tools.
A citation is not a sale. Our guide on turning AI visibility into qualified enquiries explains why accurate representation and a clear next step matter more than citation counts.
The AEO self-check.
Tick each statement that is true for your company today. You get a score out of 100 and the step to work on first. Nothing you tick is sent anywhere.
AEO readiness: 0 / 100
Tick what is true today. If you are not sure, leave it unticked.
The check points to the earliest step with a gap, because there is little value in chasing mentions while a firewall blocks the crawlers.
Questions leaders ask.
What is answer engine optimisation in simple terms?
Answer engine optimisation (AEO) is the work of getting your company named, described correctly and cited when a buyer asks an AI search tool a question.
Is AEO the same as SEO?
It is built on SEO. Google says optimising for generative AI search is still SEO and that its AI features need no special optimisations. What changes is the goal: being named and quoted inside an answer, across several engines.
What is the difference between AEO and GEO?
In everyday use, very little. AEO is short for answer engine optimisation and GEO for generative engine optimisation, a term from a 2023 research paper. Both describe making a company visible in AI-written answers.
How do I start answer engine optimisation?
Write down 20 to 30 questions your buyers ask, run them through ChatGPT, Google and Perplexity, and record who is named and which sources are cited. That shows where you are missing and why.
How do you measure AEO?
Re-run a fixed list of buyer questions each month and record the share of answers that name you. Add impressions from Search Console’s Search Generative AI performance report, citations from Bing Webmaster Tools, visits in the AI Assistant channel of Google Analytics 4, and enquiries from your CRM.
How long does AEO take to show results?
No engine publishes a timetable. Technical fixes register quickly: OpenAI says a robots.txt change can take about 24 hours to register in its systems. New pages and third-party mentions take longer, so judge progress monthly against the same question list.
If you would like help with any of the five steps, see how we work as a GEO and AEO agency.
Sources: Google Search Central, optimising for generative AI features; Google Search Central, AI features and your website; OpenAI, crawlers; OpenAI, web search tool; Perplexity, crawlers; Bing Webmaster Blog, 10 Feb 2026; Aggarwal et al., “GEO: Generative Engine Optimization”, 16 Nov 2023; G2, 6 Oct 2025; Pew Research Center, 22 Jul 2025; Conductor, AEO/GEO benchmarks report; Search Engine Journal, 30 Jan 2026; Ahrefs, 26 May 2025; Ahrefs, 4 Dec 2025; Google Search Central Blog, 3 Jun 2026; Google Analytics Help, default channel group. All accessed 12 November 2026.