AI Visibility Audit

Free guide · updated 1 Oct 2026

How to get your SaaS recommended by AI assistants

A practical checklist for small SaaS teams. No tricks: AI assistants recommend products they can read, understand and find corroborated elsewhere. Most of this takes an afternoon.

Short version:
  1. Write down 10–20 questions your buyers ask, and run them yourself in each assistant.
  2. Make sure the assistants' search crawlers aren't blocked (robots.txt, firewall, CDN bot settings).
  3. Say plainly, in text on your homepage: what it is, who it's for, and how it's different.
  4. Publish answer-shaped pages: pricing, FAQ, "X vs Y", "alternatives to Y".
  5. Get listed or mentioned on the third-party pages the assistants cite for your category.
  6. Re-run your questions every month and track the changes.

Why AI answers differ from Google rankings

When an assistant answers "what's the best tool for X?", it combines what the model already learned during training with pages it retrieves at answer time (when web search is on). That gives you two levers: be retrievable (crawlable, clearly worded pages) and be corroborated (other sites describing you consistently). Ranking well in Google helps, but it isn't the same thing. A page can rank and still be skipped if it never states plainly what the product is.

Answers also vary by user, location, time, model version and settings. Treat any single answer as a snapshot, not a ranking.

1. Test the buyer questions yourself (free)

Before changing anything, record a baseline. Write 10–20 prompts in your buyers' words, mixing:

Run each one in ChatGPT, Claude, Perplexity and Gemini (use a fresh chat with no custom instructions, and note whether web search was on). For each answer, record: were you mentioned, in what position, how you were described, which competitors appeared, and which sources were cited. A spreadsheet is enough.

Do this by hand in the apps. Their consumer terms generally don't allow automated scraping of answers; for automated testing, use the providers' official APIs.

2. Make sure AI crawlers can read your site

Each provider documents separate user agents for training and for search/answers. You can block training and still allow search. Check your robots.txt for these names:

ProviderSearch / answer-time agentsTraining agentOfficial docs
OpenAIOAI-SearchBot (ChatGPT search); ChatGPT-User (user-triggered fetches)GPTBotOpenAI crawlers
AnthropicClaude-SearchBot; Claude-UserClaudeBotAnthropic help center
PerplexityPerplexityBot (search results); Perplexity-User (user actions)(states PerplexityBot isn't used for foundation-model training)Perplexity crawlers

According to OpenAI's documentation, sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers (apart from possible navigational links), and robots.txt changes take about 24 hours to take effect. Example that allows search but opts out of training:

User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

Also check below robots.txt. Firewalls, bot-protection, JavaScript challenges and CDN settings can block crawlers even when robots.txt allows them. For example, Cloudflare announced in July 2025 that new domains would block AI crawlers by default unless the owner chose otherwise (Cloudflare press release). If you're on Cloudflare, review its AI crawler settings for your domain.

Check your content renders without JavaScript. Open your homepage and pricing page with JavaScript disabled (or curl them). If the product description isn't in the HTML, some crawlers may not see it.

3. State plainly what you are

Assistants need to match your product to a question. Put one plain-text paragraph near the top of your homepage that covers:

Then use the same wording on your profiles elsewhere. Inconsistent descriptions (old pricing on a directory, a different category on a review site) are a common reason answers describe a product wrongly.

4. Publish answer-shaped pages

Buyer questions are often comparative. Honest, specific pages give assistants something to retrieve and cite:

Structured data (Organization, SoftwareApplication, FAQPage in JSON-LD) makes facts like name, price and category machine-readable. Treat it as clarity, not a guaranteed boost.

5. Get mentioned where assistants look

When your step-1 answers show citations, list the cited domains. Typically they're review sites, "best X tools" articles, directories, community threads and documentation. Then:

What about llms.txt?

llms.txt is a proposed convention for a markdown summary of your site. As of October 2026, Google has said it doesn't use llms.txt for Search or its AI features, and we're not aware of OpenAI or Anthropic confirming that their assistants read third-party llms.txt files. It's cheap to add, but don't expect it to move results on its own. Steps 2–5 matter more.

6. Re-test monthly

Re-run the same prompts on the same assistants each month and compare: mentions, position, description accuracy, and which sources are cited. Changes to third-party pages and crawler access can take weeks to show up in answers.

Want this done for you?

Our $99 AI Visibility Audit runs 20 buyer prompts through the major assistants on your confirmed list, maps who gets recommended instead of you, and gives you ready-to-paste fixes. Or request a free mini check.

Sources

Independent guide; not affiliated with OpenAI, Anthropic, Google or Perplexity. Product names are trademarks of their owners.