AEO best practices: what actually gets a brand cited in AI answers

What are the AEO best practices that matter most?

Six practices do most of the work: sound SEO foundations, open crawler access, answer-first writing, genuine off-site mentions, regular refreshes and steady measurement. Everything else is detail inside one of those six.

The table below sets out each practice, why it matters and where the evidence comes from. We ordered it the way we run engagements: access and foundations before content, content before mentions, and measurement from day one so you can see what moved.

Why should AEO best practices start with SEO foundations?

Because the engines say so. Google's AI optimization guide, updated in July 2026, states that optimizing for generative AI search is still SEO, and its AI features documentation says a page only needs to be indexed and eligible to show a snippet, with no additional requirements.

That does not mean organic rank and AI citations are the same thing. Ahrefs found that 38% of AI Overview citations come from pages in the organic top 10, down from 76% a year earlier, so pages further down the results can still be quoted. But a page that is not indexed, renders only in JavaScript or carries a noindex tag cannot be quoted at all.

In practice the foundation checks are familiar: server-rendered HTML for the main content, a clean XML sitemap, canonical tags that point where you expect, internal links to every priority page, and no accidental noindex. Bing matters too, because Bingbot does the crawling for Copilot, so verify the site in Bing Webmaster Tools and submit URLs through IndexNow when pages change.

How should you give AI crawlers access to your pages?

Allow the crawlers that power search answers and make a separate, deliberate choice about the ones that collect training data. Mixing the two up is one of the most common and most costly mistakes we see in audits.

OpenAI's documentation separates OAI-SearchBot, which surfaces sites in ChatGPT search, from GPTBot, which collects training data; blocking GPTBot only opts you out of training. Perplexity's documentation says PerplexityBot surfaces sites in results and is not used for training. Anthropic lists Claude-SearchBot for search quality and ClaudeBot for training. Google-Extended is a control token for Gemini training and grounding that does not affect inclusion in Google Search, and AI Overviews use Googlebot.

Then check the layer above robots.txt. Cloudflare changed its default in July 2025 to block AI crawlers unless they pay, and its managed robots.txt can add disallow rules on your behalf. Read your server logs or CDN logs to confirm the search crawlers are actually fetching pages, not just permitted on paper.

• Keep allowed: Googlebot, Bingbot, OAI-SearchBot, PerplexityBot, Claude-SearchBot, Applebot.
• Decide on training: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, CCBot.
• Verify in logs: user agents fetching priority URLs with HTTP 200 responses.
• Re-check after any CDN, WAF or hosting change.

How do you write answer-first content that engines can quote?

Phrase headings as the questions buyers actually ask and answer each one in the first one or two sentences beneath it. Detail, caveats and examples follow the answer rather than lead up to it.

Google describes query fan-out: AI Overviews and AI Mode may run several related searches across subtopics before composing a response. A page whose sections each answer one clear sub-question gives the engine more chances to find a passage that fits. Google also says there is no requirement to break content into tiny pieces, so this is about clarity, not chopping pages into fragments.

Good answer-first pages share a few traits. They define terms in plain language, put numbers and named sources in the same sentence, use tables for comparisons and numbered lists for processes, carry a visible last-updated date and close with an FAQ block whose answers stand alone. They also state who the page is for, which helps an engine match it to the right prompt.

Which off-site signals carry the most weight for AI answers?

Mentions of your brand on sites the engines already trust carry more weight than any on-page tweak. In an Ahrefs study of 75,000 brands, web mentions correlated 0.664 with AI Overview visibility, against 0.218 for backlinks. That is correlation, not proof of cause, but it points the same way as the citation data.

Peec AI's analysis of 30 million sources lists Reddit, YouTube, LinkedIn, Wikipedia, Forbes, G2 and Yelp among the most-cited domains. Which of those matter for you depends on your category: G2 and Capterra for software, Yelp and Google Business Profile for local services, trade press and LinkedIn for professional services.

The best practice here is to be genuinely present: keep review profiles current, answer real questions in communities where your buyers ask them, publish original data that journalists and analysts want to cite, and make sure your Organization schema uses sameAs links to the profiles you control. Google's guide warns against inauthentic mentions, and Google's spam policies now say they apply to generative AI responses too.

Do schema markup and llms.txt help you get cited?

Yes, but as hygiene rather than levers. Google states that structured data is not required for generative AI search and that there is no special schema.org markup to add, while still recommending it for general SEO. Whatever markup you use must match the visible content.

llms.txt was proposed by Jeremy Howard of Answer.AI in September 2024. Google's John Mueller has said none of the AI services have said they use it, and no major answer engine has confirmed reading it. It costs little to publish and can help coding agents read developer docs, so we add it where it fits, but we never sell it as a route into ChatGPT or AI Overviews.

How often should you refresh pages for AI search?

Review priority pages at least quarterly and update them whenever a fact, price or product detail changes. Freshness appears to matter more in AI answers than in classic results.

Ahrefs analyzed about 17 million citations and found AI assistants cite pages that are 25.7% fresher than organic results, with ChatGPT the most freshness-leaning and AI Overviews slightly older than organic. A meaningful refresh means new facts, current examples and a corrected date, not a changed timestamp on an unchanged page. Keep a change log so you can match refreshes to movements in your prompt-set results.

How do you know whether your AEO best practices are working?

Run the same set of buyer prompts in each engine every month and record whether your brand is mentioned, cited and described correctly. Pair that with the first-party reports the engines now provide.

Google added generative AI performance reports to Search Console in June 2026, and AI feature traffic is counted under the Web search type. Bing Webmaster Tools AI Performance shows Copilot citations, cited pages, grounding queries and, since June 2026, citation share. In GA4, watch referral sources such as chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com.

The payoff can be real. Seer Interactive found that brands cited in an AI Overview earned 35% more organic clicks than those that were not. That is why we report citation share and AI referrals rather than isolated screenshots of a single good answer.

Which AEO practices should you avoid?

Avoid anything that manufactures signals instead of earning them. Short-term tricks carry legal and search risk, and answer engines tend to change faster than the tricks can keep up.

• Buying or faking reviews, which the FTC's Consumer Reviews and Testimonials Rule prohibits in the US.
• Seeding forum threads or posting undisclosed promotional answers on Reddit.
• Publishing scaled best-of lists that rank your own brand first with no clear method.
• Blocking crawlers wholesale at the CDN, then wondering why ChatGPT never cites you.
• Paying anyone who promises a placement; no one can promise a citation.