How do ChatGPT, Perplexity and Google AI Overviews choose sources?
If you want to know how to get your SaaS recommended by ChatGPT, start with a plain fact. No vendor publishes a ranking formula for AI answers. What they do publish is how their crawlers work and what makes a page eligible.
ChatGPT search relies on a crawler called OAI-SearchBot. OpenAI says this bot surfaces websites in ChatGPT's search features and is not used to train models. Its crawler documentation also says sites that block it won't appear in ChatGPT search answers.
Perplexity works in a similar way. Its crawler documentation says PerplexityBot exists to surface and link websites in Perplexity search results, and that it respects robots.txt. That is the file on your site that tells crawlers what they may visit.
Google takes a different line. Its guidance on AI features says a page must be indexed and eligible to show in Search with a snippet. It says there are no extra requirements or special optimizations for AI Overviews or AI Mode.
So the base layer is the same everywhere. The tool must be able to fetch your page, understand it, and find it relevant to the question. Past that, each system weighs sources in ways it doesn't disclose.
Can the AI crawlers reach your site?
Check crawler access first, because it is the one step that can rule you out on its own. Many sites block AI bots by accident, often through a security plugin, a CDN (content delivery network) setting, or an old robots.txt rule.
Each vendor runs more than one bot, and they do different jobs. OpenAI separates OAI-SearchBot, which serves search, from GPTBot, which collects data to train models. You can allow the first and block the second. Google offers a separate control, Google-Extended, for use of your content in its other AI systems.
- Open yoursite.com/robots.txt and look for rules that name OAI-SearchBot, PerplexityBot or Googlebot, or that block all bots.
- Ask whoever runs your CDN or firewall whether bot-blocking rules cover these crawlers.
- Decide on training bots, such as GPTBot, as a separate question from search bots.
- Confirm your key pages are indexed in Google Search Console. Google's AI features draw from the same index.
- Allow for a delay. OpenAI says robots.txt changes take about 24 hours to take effect for its crawler.
Write product pages an AI can quote
AI answers lift short, factual statements. A page that says "the modern platform for growth" gives a model nothing to repeat. A page that says "invoice software for US dental clinics with 1 to 20 staff" does.
Write each key page so a stranger could summarize it in one sentence. Say what the product does, who it is for, what it connects to, and roughly what it costs. Put those facts in text, not only in images or demo videos.
Answer the questions buyers ask in their own words. For example, a buyer might ask which tool fits a 12-person agency that already uses HubSpot. If a page answers that plainly, a model has something to cite.
- A one-line description of the product in the first paragraph of the home page.
- A page per main use case or buyer type, each with its own clear heading.
- Public pricing, or at least a clear statement of how pricing works.
- A list of integrations as text, with each one named.
- Fair comparison pages that state where you fit and where you don't.
- An FAQ with short, self-contained answers.
Does structured data help AI search?
Structured data is code on a page that labels facts, such as a product name, price or FAQ, in a standard format called schema.org. Google says it uses this markup to understand pages. Its introduction to structured data describes the details.
Be clear about what it won't do. Google's AI features guidance says you don't need special schema.org markup, new machine-readable files, or AI text files to appear in AI Overviews. Structured data makes your facts easier to parse, but it is not a switch that turns on AI mentions.
Add it where it is cheap and accurate. Organization, Product or SoftwareApplication, and FAQ markup cover most SaaS sites. Make sure the markup matches what the page says, since conflicting facts help no one.
How do third-party sites shape what AI says about you?
AI tools rarely rely on your site alone. When a buyer asks for the best tool in a category, the answer often draws on review sites, directories, comparison articles and forum threads. How others describe you matters as much as how you describe yourself.
Aim for consistency. If your site says you serve mid-market finance teams but your directory listings say "small business accounting," a model may repeat the older, vaguer version. Pick one short description and use it everywhere you control.
- Update your listings on the main review and software directory sites in your category.
- Ask happy customers to leave honest reviews that mention their use case.
- Correct outdated facts in partner pages, marketplace listings and old press coverage.
- Publish useful material others will cite, such as a benchmark method or a detailed guide.
- Show up where your buyers discuss tools, as a helpful member rather than an advertiser.
- Don't pay for fake reviews or spam forums. It can harm your reputation and break platform rules.
How do you measure AI visibility for B2B SaaS?
Measure by asking the questions your buyers ask. Write down 20 to 40 real prompts from sales calls and support tickets. Include category questions, comparison questions and "what should I use for" questions.
Run those prompts in ChatGPT, Perplexity and Google each month. Note whether you appear, how you are described, and which sources are cited. Save the answers, because they change from run to run and over time.
Watch referral traffic too. OpenAI's publishers and developers FAQ says sites that allow OAI-SearchBot can track ChatGPT referrals in analytics tools, and ChatGPT adds utm_source=chatgpt.com to links it refers. Filter your analytics by that tag.
For example, a hypothetical team might track 30 prompts across 3 tools, or 90 checks a month. If they appear in 9 answers in March and 18 in June, that is a trend worth acting on. One answer, good or bad, is not.
What is generative engine optimization, and what can't it promise?
Generative engine optimization, or GEO, is the practice of making your brand easy for AI search tools to find, understand and cite. It overlaps heavily with good SEO, clear writing and honest marketing.
Be wary of anyone who promises a spot in ChatGPT answers. The vendors don't sell one in their organic results, and answers vary by user, wording and time. You can raise your odds. You can't guarantee the result.
The work that helps is also work that helps human buyers. Clear pages, open pricing, real reviews and consistent facts pay off whether a person or a model reads them.
Where to start this month
Start with the checks that can rule you out, then work outward. Most teams can finish the first 3 steps below in a week.
- Check robots.txt and firewall rules for OAI-SearchBot, PerplexityBot and Googlebot.
- Rewrite the first paragraph of your home page and top 5 product pages as plain facts.
- Add accurate Organization, Product and FAQ structured data.
- Align your descriptions on your top 5 directory and review listings.
- Build your list of buyer prompts and record a baseline.
- Repeat the prompt check monthly and compare.
- If you want help with an audit and a plan, Valley AI's AI visibility service covers this work for SaaS teams.