What AI search optimization means
AI search optimization means making sure AI features and assistants can find your business, read your pages and describe you accurately. It covers Google's AI Overviews and AI Mode, along with tools such as ChatGPT search and Perplexity.
It isn't a separate discipline with its own secret rules. Think of it as SEO plus a few extra checks. The vendor documents below are clearer about this than most of the marketing written about them.
What stays the same
The foundations stay the same: your pages must be crawlable and indexable, and the content must be useful to people. Google says plainly that "there are no additional requirements" to appear in AI Overviews or AI Mode. To show as a supporting link, a page must be indexed and eligible to appear in Search with a snippet.
The content advice doesn't change either. Google's helpful content guidance asks whether a page offers original information, shows first-hand experience and makes clear who wrote it and why. So the same work pays off in both places:
- A clear page for each service, so there's a specific page to find.
- Direct answers to the questions customers really ask.
- Accurate business details, with a named person behind the content.
- Real experience: your own work, your own examples, your own limits.
What's different
Three things differ: how answers are put together, which crawlers you have to think about, and how you measure the results.
Answers are written, not just listed
Google says AI Mode and AI Overviews may use "query fan-out," running related searches to gather a wider set of links. AI Overviews also don't appear on every search. They show only when Google's systems judge them useful. So a customer may read a summary of your business without ever opening your site, which means your facts need to be easy to read correctly.
We handle this on our own site by keeping our prices, hours, address and awards in one source file that pages draw from, which makes it much harder for two pages to give different answers. You don't need code for this. A shared document that you check your pages against does the same job.
Each service runs its own crawlers
Search crawlers and training crawlers are usually separate, so you can make separate decisions in robots.txt.
| Crawler | What it does | What to know |
|---|---|---|
| OAI-SearchBot (OpenAI) | Surfaces sites in ChatGPT search. | Sites that opt out won't appear in search answers, though they can still show as navigational links. |
| GPTBot (OpenAI) | Collects content that may be used to train AI models. | Blocking it signals your content shouldn't be used for training. OpenAI treats it separately from search visibility. |
| ChatGPT-User (OpenAI) | Visits pages for certain user-initiated actions. | OpenAI says robots.txt rules may not apply, since a user started the action. |
| PerplexityBot | Surfaces and links sites in Perplexity's search results. | Perplexity says it isn't used to crawl content for AI foundation models. |
| Perplexity-User | Visits a page when a user's question calls for it. | Perplexity says it generally ignores robots.txt, since a user requested the fetch. |
Details come from OpenAI's bot documentation and Perplexity's. Google has its own controls, including nosnippet, max-snippet and noindex. It also has Google-Extended, a separate control for AI training and grounding in other Google systems. Check these rules before you change anything, as they're the kind of thing that gets updated.
The practical mistake to look for is a robots.txt file that blocks every AI bot because someone copied a template. If you want to be cited, that can work against you.
Measurement is less tidy
Google includes AI Overviews and AI Mode traffic in overall Search Console data, under the "Web" search type. That's useful, but it doesn't tell you what the answer said about you. For that, you have to go and read it, which is what the test below is for.
What's overhyped
The most overhyped claims are special files, special markup and promised placement.
- llms.txt. The llms.txt site itself calls it a proposal: a Markdown file offering agents brief guidance and links. It reports that thousands of sites publish one, but it's still experimental, and it isn't a Google requirement. Google says you don't need AI text files to appear in its AI features.
- "AI schema." Google says there's no special schema.org structured data you need to add for these features. Normal structured data can still help describe your business, but it isn't a key to AI answers.
- Guaranteed placement. Nobody can guarantee placement in AI answers. Google says indexing and serving aren't guaranteed, even for a page that does everything right. If a vendor promises it, ask how they control the output. They don't.
- "AI rankings." Answers can differ by wording, tool and day, so one screenshot isn't a position. Treat it as a sample.
A simple manual test you can run
You can test AI visibility yourself with a spreadsheet, a fixed list of questions and about an hour a month. It won't give you a score. It gives you a dated record of what was said, which is what you need to spot errors and changes.
- Write 10 to 15 questions customers ask. Mix ones with your name ("What does [business] charge?") and ones without it ("Who does [service] near Burbank?").
- Choose your tools. Google Search, ChatGPT and Perplexity are a sensible start.
- Log every answer in a table. Record the date, the tool, the question, what it said, whether it was accurate and whether you were cited.
- Repeat with the same questions on a set schedule. Changing them each month makes comparison impossible.
- Act on the errors. If an answer has your hours wrong, find where that came from and correct the source page or listing.
Here's the layout. The rows below show the format only and aren't real results.
| Date | Tool | Question | What it said | Accurate? | Cited us? |
|---|---|---|---|---|---|
| YYYY-MM-DD | ChatGPT | What are [business]'s hours? | Summary or pasted text | Yes / No / Partly | Yes / No, plus the link |
| YYYY-MM-DD | Perplexity | Who does [service] near Burbank? | Summary or pasted text | Yes / No / Partly | Yes / No, plus the link |
Accuracy matters more than mentions. An answer that gets your prices wrong can do more harm than one that doesn't name you at all.
What to do first
Do the SEO basics first, then check your crawler settings, then start the manual test. The order we'd use:
- Confirm your important pages are indexed in Search Console.
- Read your robots.txt and decide bot by bot.
- Give each main service a clear page that answers real questions. Our SEO service covers this groundwork.
- Make your business details match everywhere they appear.
- Run the test and fix what's wrong. If you'd like help, see our AI search optimization service.
Sources
All pages checked on 2026-11-03.
- Google Search Central guide to AI features, on requirements, controls and Search Console reporting.
- Google Search Central guide to helpful content, on people-first content and authorship.
- OpenAI crawler documentation, on OAI-SearchBot, GPTBot and ChatGPT-User.
- Perplexity crawler documentation, on PerplexityBot and Perplexity-User.
- llmstxt.org, the llms.txt proposal.
Written by Romik Hacobian, founder of Media City Design, a Burbank web design studio since 2021.