Answer Engine Optimization Guide 2026: Best Practices and Costs

In 2026, the answer engines’ own documentation boils AEO best practice down to five moves. Let their search crawlers in. Get indexed in Google and Bing. Answer each real buyer question in the first sentence under a heading written in the buyer’s words. Back every claim with a specific number or a named source. Earn genuine mentions on other sites. Google’s May 2026 guide to optimizing for generative AI features goes further: it says optimizing for AI search “is optimizing for the search experience, and thus still SEO,” and that llms.txt files, content “chunking” and special schema are not needed for Google.

This guide works from the platforms’ own words: Google Search Central, OpenAI’s crawler and publisher docs, Perplexity’s crawler docs and Microsoft’s Bing guidance. It covers question optimization with before and after rewrites, the 2026 schema and llms.txt picture, measurement in Search Console, Bing and GA4, revenue attribution, a 90-day program, results by type of firm, and verified tool and consulting prices.

What AEO means in 2026, including Google’s own view

Answer engine optimization (AEO) is the work of getting your pages used and cited as sources when an AI system answers a question: Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Microsoft Copilot and similar tools. GEO (generative engine optimization) describes the same work. If you want the vocabulary sorted out, we compare the terms in AEO vs SEO vs GEO.

Google uses both labels and then folds them back into SEO. Its AI optimization guide says AI Overviews and AI Mode are “rooted in our core Search ranking and quality systems.” They use retrieval-augmented generation, which pulls pages from the Search index, and “query fan-out,” a set of related searches the model runs to collect more results. Google’s example: a question about a weedy lawn fans out to “best herbicides for lawns” and “how to prevent weeds in lawn.” The guide also tells you to check third-party “AEO” or “GEO” advice against Google’s guidance on evaluating SEO services.

The reason the work matters is the click. In a Pew Research Center analysis of browsing data from 900 US adults in March 2025, people clicked a traditional result on 8% of visits to Google search pages that showed an AI summary. On pages without one, they clicked on 15% of visits. Links inside the summary got clicks on 1% of visits. When fewer people click, the sources named in the answer take a bigger share of the attention that is left.

How Google, ChatGPT, Perplexity and Copilot choose and cite sources

None of these companies publishes a ranking formula. Each one does document where its answers come from, which crawler to allow, and how to opt out. Here is that documentation side by side.

Google AI Overviews and AI Mode

Google’s “AI features and your website” page sets out the requirement: to be a supporting link, a page “must be indexed and eligible to be shown in Google Search with a snippet.” There are “no additional technical requirements.” AI Overviews appear only when Google’s systems judge them “additive to classic Search,” so many queries never show one. AI Mode and AI Overviews may use different models, so they can cite different links for the same question.

There is one newer setting to check. Since August 31, 2026, every site has a Search generative AI control under Settings in Search Console. “Include” is the default. If someone switched it to “Exclude,” your links and content are removed from AI Overviews, AI Mode and Discover’s generative features within a few days. Google also extended preferred sources to AI Overviews and AI Mode in May 2026. That lets loyal readers ask Google to show your site more often.

ChatGPT search

OpenAI’s crawler documentation says plainly that “OAI-SearchBot is for search.” Allowing it is the recommended way to appear in ChatGPT search results, and robots.txt changes take about 24 hours to apply. GPTBot is a separate crawler for model training. Blocking GPTBot does not remove you from search. ChatGPT-User fetches pages when a user asks ChatGPT something and “is not used for automatic web crawling.” OpenAI’s publisher FAQ adds two useful details. ChatGPT may show the link and title of a blocked page it learned about from a third-party search provider. It also appends utm_source=chatgpt.com to referral links. For an engine-specific playbook, see how to rank on ChatGPT.

Perplexity

Perplexity’s crawler docs say PerplexityBot is “designed to surface and link websites in search results on Perplexity” and is not used to train foundation models. Perplexity-User visits pages in real time to answer a user’s question, and because a person triggered the fetch it “generally ignores robots.txt rules.” Perplexity’s help center says every answer carries numbered citations that link to the original sources. Allow PerplexityBot, and allow its published IP ranges through your firewall or CDN.

Microsoft Copilot

Microsoft’s October 2025 guide, “Optimizing Your Content for Inclusion in AI Search Answers”, was written by Krishna Madhavan of the Bing team. It is the most concrete official guidance from any engine. Bing’s index powers Copilot. Assistants “break content down, a process called parsing,” then rank the pieces and often combine several sources into one answer. A May 2026 Bing post says grounding favors “discrete, supportable facts with clear provenance” and warns that “stale facts can directly produce wrong answers.”

Where the official advice disagrees

Google says you do not need to “chunk” content or rewrite it for AI. Microsoft says assistants parse pages into pieces and rewards “self-contained phrasing that makes sense when extracted.” You can satisfy both with the same habit: write complete, specific sections for human readers, each of which still makes sense if quoted alone. Do not split one topic into dozens of thin pages. Google says creating pages for every fan-out variation “primarily to manipulate rankings or generative AI responses” violates its scaled content abuse policy.

AEO question optimization best practices, with before and after examples

Question optimization means finding the exact questions buyers ask and answering each one where an engine can lift it cleanly. The rules below come from Microsoft’s guidance and Google’s documentation. The examples show the change on the page.

1. Use the buyer’s question as the heading. Microsoft calls out vague headings such as “Learn More” and says “direct questions with clear answers mirror the way people search.” Take the wording from your Search Console queries, Bing’s grounding queries and your sales calls, not from a brainstorm.
2. Answer in the first sentence. Lead with the yes, the no, the number or the name. Explain after that.
3. Put a specific in every answer. Microsoft lists “avoid vague language” among its mistakes and asks you to anchor claims in measurable facts. Use a number, a range, a timeframe, a condition, a statute or a named source.
4. Make each section stand alone. No “as mentioned above.” Name the subject instead of starting with “it” or “this.”
5. Keep the answer in visible HTML text. Microsoft says: “Don’t hide important answers in tabs or expandable menus,” don’t rely on PDFs for core information, and don’t put key facts only in images.
6. Align the title, H1 and meta description. Microsoft wants the H1 to “match (or closely reflect) the page title,” with all three stating the same purpose.
7. Answer the follow-up on the same page. Fan-out means the engine is also searching the next question. Cover it in an H3 on the page that owns the topic instead of publishing a new thin page.

Before and after: a question about AEO itself

Question: Do I need llms.txt to appear in Google AI Overviews?

Before: “llms.txt is an exciting new standard that could transform how AI understands your website. Many experts believe it will soon become essential for any serious AEO strategy.”

After: “No. Google’s AI optimization guide says Google Search ignores llms.txt files, so adding one neither helps nor hurts visibility in AI Overviews or AI Mode. Other systems may read the file, so keeping one does no harm.”

The rewrite answers in one word, names the source and gives the condition. Every claim in it can be checked.

Before and after: a consulting firm’s pricing question

Question: How much does a pricing strategy project cost? (Template: replace the brackets with your real figures.)

Before: “Every engagement is unique, and we tailor our approach to each client’s needs. Contact us to learn more about our flexible pricing options.”

After: “Our pricing strategy projects for firms with $[X]M to $[Y]M in revenue are fixed-fee, from $[low] to $[high], and take [N] weeks. The fee covers a win/loss analysis of your last [N] proposals, a price test on one service line and a rollout plan. It rises when more than [N] service lines are in scope.”

An engine can’t quote “it depends.” A buyer asking an AI assistant what consulting projects cost gets an answer built from firms that publish a range.

Before and after: a law firm’s process question

Question: How long does probate take in [State]? (Template.)

Before: “Probate can be a long and complicated process, and every estate is different. Our experienced attorneys will guide you every step of the way.”

After: “Uncontested probate in [State] usually takes [N] to [N] months from filing to final distribution. Creditors have [N] months to file claims under [statute citation], and the estate cannot close before that window ends. A will contest or a missing beneficiary adds time.”

The statute citation is the provenance Bing describes. It also keeps the claim accurate, which matters for any firm working under attorney advertising rules.

The question-section scorecard

Score every question section on these eight pass/fail checks before you publish it.

8 of 8: publish. 6 or 7: fix the failures. 5 or fewer: rewrite the section. To score individual sentences rather than sections, use the extractability test in our 12-month AEO roadmap. It scores the 10 sentences that carry a page’s main claims, which answers “how do I measure sentence-level extractability.”

Website structure for AEO: H1, headings and FAQ strategy in 2026

One page per topic. One H1 that matches the title. H2s for the main questions, H3s for the follow-ups, and tables for anything with more than two attributes. Microsoft says lists and tables work well for how-to queries and comparisons. Google says people appreciate content “organized by paragraphs and sections, along with headings.”

FAQ strategy changed this year. Google deprecated the FAQ rich result and stopped showing it in Search on May 7, 2026. It removed the documentation on June 15. FAQPage markup no longer earns a visual feature on Google. A short FAQ block still has a job, though. It answers real follow-up questions in Q&A form, which Microsoft says assistants “can often lift directly into responses.” Keep FAQs for real questions with specific answers, and drop the padding questions nobody asks.

Which schema matters for AEO, according to the engines

The practical rule: keep accurate Organization, Article and author markup that matches the page. Don’t expect any schema type to win a citation by itself.

llms.txt in 2026: the honest status

llms.txt is a proposal by Jeremy Howard, published September 3, 2024 and revised as “v2” on August 10, 2026. It adds a Markdown file at your site root that summarizes the site and links to clean Markdown versions of key pages. Documentation platforms generate it automatically. Perplexity’s own help center points visitors to one.

For Google, the answer is settled. On June 15, 2026, Google added a note to its AI optimization guide saying llms.txt files are not used by Google Search and “won’t negatively or positively impact your visibility or rankings.” None of the OpenAI, Perplexity or Microsoft crawler and publisher pages cited above mention llms.txt as a way to get into their search results. My view: if your CMS makes one in ten minutes, fine. It should never be the first item in an AEO plan, and anyone selling it as a citation lever is ahead of the evidence.

How to measure AEO in Search Console, Bing Webmaster Tools and GA4

Google Search Console

1. Confirm eligibility. Settings, then Search generative AI. The control should read “Include my site’s links and content.”
2. Open the Generative AI performance report. It shows impressions from AI Overviews and AI Mode by page, country, device and date. It has been available to all sites since August 31, 2026. An impression counts when a link to your site appears in the AI feature.
3. Read the main Performance report correctly. Under Google’s counting rules, every link in an AI Overview shares a single position. A follow-up question in AI Mode counts as a new query.
4. Find your conversational queries. Use the Custom (regex) query filter, which takes RE2 syntax. Try ^(how|what|which|why|should|can|do|does|is)\s to isolate question queries, or (\S+\s){9,} for queries of ten words or more. Long, prompt-style queries are where AEO demand shows up first.
5. Track brand demand. The branded queries filter, available to all eligible sites since March 11, 2026, separates branded from non-branded queries. A rise in branded impressions is the most honest sign that people saw you in an answer and then searched for you.

Bing Webmaster Tools

The AI Performance report launched in public preview on February 10, 2026. It covers Copilot, AI summaries in Bing and select partners. It reports Total Citations, Average Cited Pages, grounding queries (the phrases the AI used to retrieve your content) and citations by page. A June 16, 2026 update added Intents, Topics, Citation Share (your percentage of all citations for a grounding query) and period Compare. Bing describes the data as a sample, and it has no click data. It is still the only first-party citation count any engine gives you.

GA4

On May 13, 2026, Google Analytics added an AI Assistant channel to the Default Channel Group. Matching referrals get the medium “ai-assistant” and the campaign “(ai-assistant).” The channel definitions page lists sources “like ChatGPT, Gemini, Deepseek, Copilot, or Grok.” It excludes Google’s AI Overviews and AI Mode, which stay in Organic Search. Perplexity is not named in that list, so build a custom channel group as a backstop, with a referrer regex such as chatgpt\.com|perplexity\.ai|copilot\.microsoft\.com|gemini\.google\.com|claude\.ai, and compare the two. Our AI search analytics guide covers the reporting layer in more depth.

How do I attribute revenue to answer engine optimization?

Attribute AEO revenue in three layers and add them up without double counting. The first layer is GA4 conversions in the AI Assistant channel. The second is self-reported source: add “ChatGPT, Perplexity or another AI assistant” as an option on your intake form’s “How did you first hear about us?” question. The third is influence: branded query growth in Search Console and citation counts in Bing, reported next to revenue but never added to it.

Worked example, with hypothetical numbers so you can see the math. An accounting firm gets 120 new inquiries in a quarter.

Compare that figure with the quarter’s AEO cost and you have a defensible ROI line for a CFO. Self-reporting undercounts, because many buyers ask an assistant and then simply search your name. That is why the branded-demand layer sits beside the number, not inside it.

A 90-day answer engine optimization program

Ninety days is long enough to fix access, measure a baseline, rewrite your highest-value pages and run the first re-measurement. It is not long enough to judge off-site authority. That is quarter two and beyond, covered in the 12-month roadmap.

Two notes on running this. First, each engine changes models on its own schedule, which is the worry behind “each AI model has different update cycles.” So keep the prompt set fixed, re-run it monthly, run each prompt more than once, and log the date of each run. A shift that appears in one engine only is usually that engine changing, not your work. Second, you don’t need a separate AEO roadmap. Google says this work “is still SEO,” so add three columns to your existing SEO backlog: target question, engine evidence (cited or not) and scorecard result.

What results to expect from AEO, by type of firm

Expect foundations in the first 90 days and specific-question citations before broad recommendations. Ahrefs’ study of 75,000 brands (December 2025) explains why. Branded web mentions correlated with AI visibility at 0.66 to 0.71, and YouTube mentions at about 0.737. The number of pages on a site correlated at about 0.194. The study’s own summary: “The big brands always come out on top.” These are correlations, not proof of cause. They still point to the same lesson: publishing more pages is not what gets a smaller firm cited.

For vertical playbooks, see how estate planning attorneys get cited by ChatGPT and how accounting firms get cited by AI search.

AEO tools and what they cost in 2026

Every tool below runs a set of prompts through AI engines and records whether your brand is mentioned or cited. They differ in how many prompts you get, which engines they cover and what they add on top. Prices are from each vendor’s pricing page in October 2026, monthly billing unless noted.

Profound or Evertune for automating AEO content?

On published plan details alone, Evertune is the one with a priced content quota: its Pro plan includes a Content Agent at 25 articles a month for $800 a month billed annually. Profound’s AI Marketer and reusable agents also research and write content, but pricing requires a demo, so you can’t compare costs before a sales call. Pros for Evertune: a public price and a fixed article quota. Cons: the annual commitment. Pros for Profound: up to 9 answer engines and agent workflows. Cons: no public price for paid plans. For a professional-service firm, my opinion is that neither should write your answers unsupervised. Google’s guidance on generative AI content says to “manually factcheck and review all AI-generated content,” metadata included.

Setup mistakes that waste the first month

• Tracking prompts that only contain your brand name. You will look visible on prompts nobody types.
• Running each prompt once. Answers vary between runs, so one run is an anecdote.
• Paying for engines your buyers don’t use. Ask new clients which assistant they use before choosing models.
• Skipping the competitor set. Visibility means little until you see who gets cited instead of you.
• Treating any tool’s score as Google data. Google says no third-party tool “has access to our internal ranking or AI systems.”

On budget: a first-year stack of a $95 to $245 tracker plus analyst time is a small test, not a platform decision. Our roadmap includes a worked example of justifying a 10% shift from the SEO budget to a CFO.

What AEO consulting includes and what it costs

A real AEO engagement covers five things: crawler and indexing access across all four engines, a measured prompt baseline, a question map, rewrites of priority pages, and an off-site mention plan. A consultant who also handles content production adds the writing, with a fact-check step for every claim. Here are the published price points I could verify:

The MarketerHire figures are a marketplace’s published ranges, not a survey, and enterprise programs sit at the top of them. Vet any provider the way Google suggests in its third-party SEO guidance. Be wary of guaranteed rankings or citations, and of claims to “internal” Google data. A “no fluff” engagement shows you the baseline, the pages it will change and the re-measurement date before you sign. That is how our Rank on AI engagement is scoped. To see whether AEO is the right lever for your firm this quarter, you can book a consultation.

Common AEO mistakes in 2026

• Blocking the wrong bot. Disallowing OAI-SearchBot to stop “AI training” removes you from ChatGPT search. GPTBot is the training crawler.
• Leaving a firewall rule in place. Google lists crawling allowed “by any CDN or hosting infrastructure” as a basic. Perplexity and OpenAI both publish IP ranges for allow-lists.
• Building fan-out pages. One thin page per prompt variation runs into Google’s scaled content abuse policy.
• Buying mentions. Google calls inauthentic mentions less helpful “than it might seem” and says its AI features rely on its spam systems.
• Letting facts go stale. Bing warns that stale facts produce wrong answers. Date your figures and refresh them.
• Measuring only sessions. Many citations produce no click. Track impressions, citations and branded demand next to traffic.

An AI-assisted writing workflow that stays trustworthy

1. A person builds the source pack first: primary documents, your own data and pricing, and the exact buyer questions.
2. The AI drafts an outline and a first pass from that pack only.
3. The subject expert adds what only the firm knows: fees, timelines, edge cases and opinions.
4. An editor fact-checks every number, date and quote against the source pack, plus the title, meta description and alt text, as Google’s guidance requires.
5. The editor reads the draft aloud and cuts any sentence that would fit on a competitor’s site unchanged.

Frequently asked questions

What is answer engine optimization (AEO) in 2026?

AEO is the work of getting your pages used and cited as sources when AI systems such as Google AI Overviews, AI Mode, ChatGPT search, Perplexity and Copilot answer a question. Google's May 2026 guide says that optimizing for its generative AI features is still SEO: pages must be indexed and eligible for a snippet, and no special markup or AI text file is required.

What are the best practices for answer engine optimization in 2026?

Allow the search crawlers (Googlebot, Bingbot, OAI-SearchBot, PerplexityBot). Use buyers' real questions as headings and answer each one in the first sentence. Put a number, condition or named source in every answer. Keep answers in visible HTML, not tabs or PDFs. Earn genuine third-party mentions. Measure in Search Console, Bing Webmaster Tools and GA4.

What results should a management consultancy realistically expect from AEO within the first six months?

Mostly foundations: crawler access fixed, a measured prompt baseline, rewritten priority pages, and citations on narrow questions where the firm publishes specifics, such as project timelines or fee ranges. Being named for broad prompts like best consulting firms usually takes longer, because those answers lean toward brands with many third-party mentions. No one can guarantee a citation.

How do I attribute revenue to answer engine optimization?

Use three layers. Count GA4 conversions in the AI Assistant channel, which Google added on May 13, 2026. Add an AI assistant option to your intake form's source question and remove leads counted twice. Multiply unique AI-sourced leads by close rate and average fee. Report branded search growth and Bing citation counts alongside that figure as influence, but do not add them to revenue.

Do I need llms.txt or FAQ schema for AEO?

Not for Google. Google's AI optimization guide says Google Search ignores llms.txt, and that no special schema is needed for AI Overviews or AI Mode. Google stopped showing FAQ rich results on May 7, 2026. Accurate structured data that matches the page is still worth keeping for SEO, and a short FAQ that answers real questions still helps readers.

What does answer engine optimization consulting cost in 2026?

Published ranges vary widely. MarketerHire lists fractional AEO consultants at $3,000 to $15,000 per month, audit and build projects at $8,000 to $25,000, and agency retainers at $10,000 to $35,000 or more per month. AEO Engine's managed Growth plan is $1,597 per month on a 90-day initial term. Tracking tools start around $29 to $99 per month.

How do I build an AEO roadmap for the next 12 months?

Spend the first quarter on crawler access, analytics setup and a baseline across ChatGPT, Perplexity, Copilot and AI Mode. In the second quarter, rewrite priority pages so each question is answered first, with specifics and sources. Use the third quarter to earn third-party mentions, and the fourth to re-measure, prune and repeat. Keep it inside your SEO backlog, not as a separate plan.

Engine-specific AI search playbooks

Each major AI engine sources and cites answers differently. These guides break down the mechanics engine by engine:

• How to rank on ChatGPT (Bing index + earned media)
• How to get cited by Claude (crawler access + primary data)
• How to rank on Perplexity (citation-first optimization)