How to Optimize for ChatGPT Search and Perplexity

Short answer: AI search tools retrieve pages from the live web and compose an answer with citations. To be eligible, your pages must be reachable by their crawlers, rank reasonably in the underlying search they rely on, and contain clear, specific passages that answer the question. Third-party pages that mention you count as much as your own. There are no published ranking factors, so be wary of anyone claiming a formula.

Table of Contents

1. How AI Search Finds Content
2. Step 1: Be Reachable
3. Step 2: Be Findable
4. Step 3: Be Quotable
5. Step 4: Be Corroborated
6. Step 5: Measure
7. What You Cannot Control
8. Key Takeaways

How AI Search Finds Content

When a question needs current information, these products typically:

1. Turn the question into one or more search queries.
2. Retrieve candidate pages, from their own index or a search provider.
3. Read relevant passages from those pages.
4. Write an answer and cite the sources used.

So appearing in an AI answer depends on three things: being retrievable, having a passage worth using, and being trusted enough to be selected. The exact methods are not public and they change.

Step 1: Be Reachable

This is first because everything else depends on it.

• Allow the search and user-triggered agents. For OpenAI these include OAI-SearchBot and ChatGPT-User; for Perplexity, PerplexityBot and Perplexity-User. These are separate from training crawlers (crawler guide).
• Check the robots.txt that is actually served, and whether your CDN blocks these agents at the edge (how to check).
• Put content in the HTML. Text that only appears after JavaScript runs may not be read.
• Avoid walls. Login requirements, aggressive bot challenges, and interstitials stop crawlers.
• Keep pages fast and stable.

Step 2: Be Findable

AI search starts from search. A page that cannot be found by a conventional search for the topic is unlikely to be retrieved.

• Get indexed in the major search engines. Submit a sitemap to both Google and Bing; several AI products draw on more than one index.
• Cover the questions people ask, phrased as they ask them.
• Use descriptive titles and headings.
• Fix technical basics: consistent canonical URLs (trailing slash bug), honest sitemap dates (lastmod).
• Build authority. Links and mentions help you rank in the search that feeds the assistant.

Step 3: Be Quotable

Once a page is retrieved, a passage has to be worth using.

• Lead with the answer at the top of the page and under each heading.
• Write self-contained paragraphs that make sense when lifted out.
• Be specific: numbers, names, steps, limits.
• Use tables for comparisons and numbered lists for procedures.
• Date the content and keep it current. AI search favours recent information for many queries.
• Include an FAQ with direct answers.
• Add structured data so facts are unambiguous (schema markup).

More detail in how to write content AI assistants cite.

Step 4: Be Corroborated

Look at the sources an assistant cites for a category question such as "best tools for X". They are often not vendors' own sites. They are comparison articles, review platforms, community threads, and publications.

That means your presence on other people's pages matters at least as much as your own content.

• Find the cited sources for your key questions and note which appear repeatedly.
• Get included where it is legitimate: directories, review sites, comparison articles, relevant communities.
• Earn independent coverage by having something worth covering: data, a strong opinion, a useful free tool.
• Keep facts consistent across every place you are described. Conflicting prices or descriptions weaken trust.
• Encourage genuine reviews.

Do not fabricate reviews or seed forums with disguised promotion. It is dishonest, communities remove it, and it damages the trust you are trying to build.

Step 5: Measure

• Mentions and citations. Run a fixed prompt set monthly (AI share of voice).
• Which of your pages are cited. Those are the formats that work.
• Crawler activity. Are the search agents fetching your pages?
• Referral traffic from assistants, tracked as its own channel (how).
• What those visitors do. Sessions from AI referrals often land deep in the site and look for one specific thing. Watch a few and make sure cited pages have a clear next step.

DeepSync's AI visibility feature, on Business and Enterprise, reports AI crawler activity, referral traffic and citations from answer engines, and brand mentions for chosen topics.

What You Cannot Control

• Which sources are chosen for a given answer.
• How you are described.
• Variation between identical questions.
• Model and product changes that shift results overnight.
• Personalisation based on a user's history.

Treat AI search as a channel with real potential and limited predictability. Do the basics well, measure monthly, and do not stake your pipeline on it.

Key Takeaways

• AI search retrieves live pages; reachability comes first.
• Being findable in conventional search is a prerequisite.
• Clear, specific, self-contained passages get used.
• Third-party mentions shape answers as much as your own site.

Conclusion

There is no trick to AI search. Let the crawlers in, be findable, answer clearly, and be the kind of company others mention. Then measure and keep going.

Check whether your CDN is blocking AI crawlers.