Michael Preuss · 2026-02-16 · Updated 2026-11-22 · 5 min read

What I Mean by Answer Engine Optimization

Answer engine optimization (AEO) as of October 2026: answer the questions that decide purchases, check what AI answers say about us, and use a weekly review to decide what to improve.

Read with...

Pick an agent

Gemini and Your Claw use copy and paste.

A buyer can decide we're not right for them without ever opening our website. They can ask an answer engine what we do, whether we fit their needs and how we compare. We need to know what that answer says.

In Pew's analysis of Google searches made by 900 U.S. adults in March 2025, users clicked a traditional result on 8% of visits with an AI summary, compared with 15% without one. That's all of their searches, not only buying research. (Pew Research Center)

Clicks still matter, but they don't tell us how we're described to someone who never visits. That's the work I mean by answer engine optimization, or AEO: publish clear, supported answers to buying questions, then check where we appear and how we're represented.

Answer the buying questions

Start with the questions sales and support hear in real conversations. What does it cost at our scale? Will it work with our existing tools? What will we have to change? When is another product a better fit?

Josh Grant's question-mining guide is the cleanest operating frame I've seen: keep a living backlog of real buyer questions, grouped by intent and reviewed weekly.

Group related questions by what the buyer needs to decide. Several versions of "Are we too small for this?" can belong on one page. Ethan Smith's question-research approach at Graphite makes the same distinction: a page can answer a group of questions with the same intent.

That page needs a direct answer, the conditions that change it and evidence the buyer can check. Put the limits next to the claim. A price without its minimum commitment, or a performance result without its test conditions, leaves out information someone needs to make the decision.

Pricing, comparisons, integrations, security and getting started are useful places to begin. Improve the pages that already answer those questions before creating more. Someone should be able to understand the answer without piecing it together from several pages.

Make sure the right systems can read it

Google says the existing SEO fundamentals still apply to AI Overviews and AI Mode. Pages need to be indexed and eligible to appear with a snippet. Google doesn't require special AI files or schema for those features. Clear text, internal links, crawl access and structured data that matches the visible content still matter. (Google Search Central)

Search, model training and visits requested by a user are different uses of the site. As of October 2026, OpenAI and Anthropic document separate bots for each. (OpenAI, Anthropic)

PurposeOpenAIAnthropic
Search discoveryOAI-SearchBotClaude-SearchBot
Model trainingGPTBotClaudeBot
User-requested visitsChatGPT-UserClaude-User

Google handles it differently: AI Overviews and AI Mode follow Googlebot's rules for Search, and Google-Extended is the separate control for training and grounding in some of Google's other systems. (Google Search Central)

Decide which uses we want to allow, then check both robots.txt and the rules in our CDN or web application firewall. Check the provider's documented bot identities and IP ranges when configuring access. Perplexity, for example, recommends checking both the user agent and the source IP. (Perplexity)

User-requested visits don't all follow the same rules. OpenAI says robots.txt may not apply to ChatGPT-User, and Perplexity says its user-requested fetches generally ignore those rules. Anthropic says its bots honour them. Read the rules for the service we're configuring. (OpenAI, Perplexity, Anthropic)

Getting access right makes the content available. It doesn't guarantee that an answer engine will cite it.

Look at what gets cited

Our own website is only part of the work. The sources an engine cites can include conversations where other people explain, compare or question the product.

At Blaxel, I built a citation-review tool around Profound's answer-engine data. It identified frequently cited Reddit threads, topics where Blaxel was less visible and threads gaining citation share. Those signals became a weekly queue for human review before any participation. The workflow is in AI Marketing Systems.

That gives the human review a specific job: inspect the source and decide whether we have a useful contribution to make.

Keep a weekly review

Use a consistent set of buyer questions, include variations, and repeat them across the answer engines our customers use. Keep the questions, dates, answers and cited URLs. One answer only shows what happened in that run. Graphite's measurement approach tracks appearances across engines, question variants and repeated runs.

Review three things. Josh Grant's 2026 AEO guide frames them as visibility, comprehension and conversion:

  • Where we appear. Which relevant answers mention us, which cite our pages, and which leave us out?
  • What the answer says. Does it get our capabilities, pricing and fit right? Does it preserve the conditions attached to our claims?
  • What buyers do. Which identifiable AI referrals lead to inquiries, qualified opportunities and revenue?

Keep citation changes separate from the commercial results we can actually attribute. Google includes traffic from AI Overviews and AI Mode in Search Console's overall Web search data. (Google Search Central)

Start with 20 real buying questions and improve the answers to the five that matter most. Each week, review the results, assign the next fixes and inspect the sources that keep appearing. The next week's work should come from what the review shows.

FAQ

Questions this piece answers

What is answer engine optimization?

AEO means publishing clear, supported answers to buying questions, then checking where we appear and how we're represented. The work includes our own website and the other sources answer engines cite.

Does AEO replace SEO?

No. SEO fundamentals still apply. Google says AI Overviews and AI Mode do not require special AI files or schema. Crawl access, internal links, clear content and accurate structured data still matter.

What should we publish first?

Start with real questions from sales and support about cost, fit, integrations and getting started. Group questions with the same intent, improve existing pages where possible, and put the answer, its limits and its evidence together.

How do we measure progress?

Review whether we appear, whether the answer is accurate, and what identifiable AI referrals lead to. Repeat questions across engines and runs. Keep citation changes separate from commercial outcomes we can actually attribute.

Why review sources outside our own website?

Answer engines can cite conversations where other people explain or compare the product. At Blaxel, I built a tool using Profound data to put relevant Reddit sources into a weekly queue for human review before any participation.