When a buyer asks ChatGPT, Perplexity, Claude or Google's AI for "the best tool for X", the assistant searches the web, reads what it finds and names the products that are described clearly and recommended elsewhere. For a SaaS company that means three things: let the AI search crawlers read your marketing site and docs, publish pages that answer buyers' questions in plain text (pricing, integrations, comparisons, use cases), and earn genuine reviews and mentions on the sites assistants cite. This guide covers each, with a 30-day plan.
Key takeaways
- Software buyers ask AI five kinds of question: best-for, versus, alternatives, integrations and pricing. Your site should have a page that answers each.
- Assistants only quote what their crawlers can fetch. Pricing hidden behind "Contact sales", docs behind a login and pages rendered only in the browser all count as missing.
- G2 and Capterra do not block the main AI search crawlers (checked 10 November 2026), so what reviewers write there can end up in answers.
- Comparison and alternatives pages are the pages assistants need most and most SaaS sites lack.
- Measure by asking the assistants the same questions every month, not by waiting for a referral to show up in analytics.
What do software buyers ask AI assistants?
Buyers ask AI assistants the same questions they used to ask a colleague or a review site, and each type needs a different page to answer it.
These are illustrative examples of the kinds of questions people ask, not measured search volumes:
| Question type | Example (illustrative) | Page that answers it |
|---|---|---|
| Best-for | "Best project management tool for a 10-person agency" | Use-case page per customer type |
| Versus | "Notion vs Confluence for engineering docs" | Comparison page per main rival |
| Alternatives | "Cheaper alternatives to HubSpot for a small charity" | Alternatives page that names rivals honestly |
| Integration | "Does Xero integrate with Shopify?" | Integrations page, one entry per integration |
| Pricing | "How much does Intercom cost for 5 seats?" | Pricing page with numbers in plain text |
Each question is answered by a page, not by a brand. An assistant answering "X vs Y" looks for a page that compares X and Y. If you have not written one, it will use a review site's comparison, a rival's page or a forum thread.
Where do AI assistants get their answers about software?
From live web search, using their own crawlers and indexes, plus third-party review and listing sites; none of them publishes a ranking formula.
What each platform documents:
- ChatGPT "sometimes partners with other search providers", rewrites your question "into one or more targeted queries", and ranks results using "multiple factors"; "placement is not guaranteed". To be eligible, "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses" (OpenAI Help Center, 2026). "Each setting is independent of the others", so you can block the training crawler GPTBot and still allow OAI-SearchBot (OpenAI, 2026).
- Perplexity runs its own index, described in September 2025 as "covering hundreds of billions of webpages" (Perplexity, 2025). PerplexityBot "is designed to surface and link websites in search results on Perplexity" and "is not used to crawl content for AI foundation models" (Perplexity, 2026).
- Claude grounds answers "with content from the live web", and "every response includes citations" (Anthropic, 2026). Blocking its crawler Claude-SearchBot "may reduce your site's visibility and accuracy in user search results" (Anthropic, 2026).
- Google AI Overviews and AI Mode use the Google index. A page must be "indexed and eligible to be shown in Google Search with a snippet", and there are "no additional requirements" beyond that (Google Search Central, 2025). Google's AI features can also show "what's being said about products and services across the web, including in blogs, videos, and forum discussions" (Google Search Central, 2026). The Gemini app's grounding is controlled by the Google-Extended robots.txt token, which "does not impact a site's inclusion in Google Search" (Google Search Central, 2026).
Our guide to how AI assistants choose businesses covers each platform in depth. The practical point: your own site, the review sites and the places developers talk about tools all feed the same answers.
Which pages does a SaaS site need for AI visibility?
Seven pages, each answering one question type in plain text that a crawler can read without running JavaScript or logging in.
Google says "you don't need to write in a specific way just for generative AI search" and favours "unique expert or experienced takes" over commodity content (Google Search Central, 2026). The pages below are not AI tricks; a careful buyer wants them anyway.
- Pricing page with numbers in text. Put each plan's price, billing period, seat limits and inclusions in HTML text, not in an image or a widget that loads later. If you sell by quote only, publish a starting price or a typical range and say what moves it. Google's
SoftwareApplicationmarkup requiresnameandoffers.priceplus a rating or review (Google Search Central, 2026), but structured data "isn't required for generative AI search" (Google Search Central, 2026), so treat it as tidy-up, not foundation. - Integrations page. One section (ideally one URL) per integration, named the way buyers name it ("Xero", "Slack", "Shopify"), with what syncs, in which direction, and on which plan. "Does X integrate with Y" is the easiest question to win because few sites answer it in text.
- Comparison pages. One page per main rival: "YourProduct vs Rival". Use a table with the same rows for both products (price, limits, integrations, support, hosting region), state the rival's facts accurately, and say who each suits. Our guide to writing a comparison page that AI quotes has a template.
- Alternatives page. "YourProduct alternatives" or "Alternatives to Rival". Name real alternatives, including ones that are better for some buyers. An honest page is more quotable than one that only lists you.
- Use-case pages. One per customer type or job: "for agencies", "for accountants", "for charities". Lead with the specific problem and the features that solve it.
- Changelog. A public, dated list of what changed. It answers "does X do Y yet" with a date.
- Docs without a login. At least the getting-started guide, the API overview and the integration guides should be public, server-rendered pages. Google notes that "not all bots can run JavaScript" and that "server-side or pre-rendering is still a great idea" (Google Search Central, 2026). View the page source: if the words are not in it, crawlers may not see them.
Two checks on the site itself:
- robots.txt. Allow
OAI-SearchBot,PerplexityBot,Claude-SearchBot,Googlebotandbingbot. Blocking training crawlers (GPTBot,ClaudeBot) is a separate choice that does not remove you from OpenAI's or Anthropic's search answers. The AI visibility checklist has a copy-and-paste example and the guide to allowing ChatGPT to crawl your site covers CDN and firewall settings. - Search Console. The "Search generative AI control" governs whether a site appears in AI Overviews, AI Mode and generative AI features in Discover. Inclusion is the default and the control "doesn't affect AI training" (Google Search Console Help, 2026). Make sure nobody has switched it off.
How do G2, Capterra and other review platforms feed AI answers?
Review platforms are crawlable, structured and full of comparative language, which is what an assistant reaches for when it has to say which tool is "best".
G2. G2 says "every company can claim and maintain a free G2 Profile", that you "follow the 'Claim Profile' workflow to verify ownership", and that you can update product details, screenshots, videos and pricing but "cannot edit or remove user reviews". Paid plans "unlock enhanced customization, branding, analytics, and lead generation capabilities" (G2, 2026). So the free step is to claim the profile, make pricing and categories match your site, and ask real customers to review.
Capterra. Capterra's vendor pages were behind a bot check when we tried to open them on 10 November 2026, so we do not describe its listing process here; look for the vendor or "get listed" link on capterra.com. Its robots.txt, which we could read, is summarised below.
What their robots.txt files say about AI crawlers (both read on 10 November 2026):
| Site | AI crawler rules |
|---|---|
| g2.com (G2 robots.txt) | A group for GPTBot, ClaudeBot, Applebot-Extended, Meta-ExternalAgent, Amazonbot and CCBot repeats the general rules and adds blocks on individual review pages (/products/*/reviews/*), three-way compare pages (/compare/*-vs-*-vs-*) and the German, French, Spanish, Portuguese and Italian sections; product and category pages stay open. OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, Claude-User and Googlebot are not named, so they follow the general * rules. Bytespider, DeepSeekBot, Brightbot and a list of scraper tools are blocked from everything. The file ends with Sitemap: https://www.g2.com/llms.txt. |
| capterra.com (Capterra robots.txt) | One general * group blocks search, preview, login, PPC and similar paths. No OpenAI, Anthropic, Perplexity, Google or Common Crawl token is named, so those crawlers follow the general rules. A list of scraper tools, plus Bytespider and DeepSeekBot, is blocked from everything. The file includes a non-standard LLMS TXT: https://www.capterra.com/llms.txt/ line. |
In plain terms: the search crawlers behind ChatGPT, Perplexity, Claude and Google can read your G2 and Capterra listings, so what is on them matters. Keep pricing, category and description identical to your site; contradictions give an assistant a reason to hedge or pick someone else. The same goes for Trustpilot; see does AI use Trustpilot reviews.
Ask every customer for a review, not just the happy ones, and never pay for or write them. Google warns that "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem" (Google Search Central, 2026), and G2 says vendors cannot remove reviews, so a listing full of suspiciously similar five-star reviews is a risk you keep.
Do developer communities and forums matter for SaaS visibility?
Yes, because assistants read them, and because Google says its AI features can show what is said "in blogs, videos, and forum discussions" (Google Search Central, 2026).
For a developer-facing product, the questions people ask in public (on Stack Overflow, GitHub issues, Reddit, Hacker News and niche forums) are the questions an assistant has to answer. What helps, in order of effort:
- Answer your own support questions in public docs. Every recurring ticket becomes a docs page with the question as the heading. That page is quotable; a ticket is not.
- Keep GitHub repositories and SDK READMEs current. A README is often the first thing a crawler reads about a developer tool.
- Take part honestly where your users already are. Answer as yourself, disclose that you work on the product, and recommend a rival when it fits.
- Publish engineering posts with real detail: an outage write-up, a benchmark with the method shown, a migration guide from a rival. Google's phrase for what stands out is "a unique viewpoint" (Google Search Central, 2026).
How do I measure AI visibility for a SaaS product?
Ask the assistants your buyers' questions on a fixed schedule, record who is named and which sources are cited, and track the change.
- Write 15 to 25 questions across the five types above, with your real category, rivals, integrations and customer types filled in. Keep the wording fixed.
- Ask ChatGPT, Perplexity, Claude and Google AI Mode each question once a month, with web search on. Answers vary between runs, so ask each twice.
- Record three things per answer: whether you were named, which rivals were named, and which URLs were cited. The cited URLs are your to-do list: if G2 is cited and you are not on it, that is the gap.
- Check Search Console's Generative AI performance report, which Google recommends for measuring AI features (Google Search Central, 2026).
- Tag referrals from chatgpt.com, perplexity.ai and claude.ai as their own analytics channel, but treat them as a lagging signal; being named without a click still shapes a shortlist.
Our guide to checking whether ChatGPT recommends your business walks through the manual method.
What should a SaaS do in the first 30 days?
Fix eligibility in week one, publish the missing pages in weeks two and three, and start the review and measurement routine in week four.
Week 1: eligibility
- Check robots.txt and your CDN or firewall for blocks on the AI search crawlers listed above.
- View the source of your pricing page, one docs page and one integration page. If the words are not there, ask engineering about server-side rendering.
- Confirm the Search Console generative AI control is set to include your site.
- Claim your free G2 profile and check whether you have a Capterra listing; make pricing and category match your site.
- Write your fixed question set and run it once. This is your baseline.
Week 2: pricing, integrations, docs
- Rewrite the pricing page with plain-text numbers and inclusions, with a short FAQ under the plans.
- Build the integrations page with one entry per integration and plan availability.
- Move the getting-started guide, API overview and integration guides out from behind any login.
Week 3: comparison, alternatives and use cases
- Publish one comparison page against the rival named most often in the baseline run.
- Publish an alternatives page that names real alternatives.
- Publish two use-case pages for the customer types that appeared in the "best for" answers.
- Start a public, dated changelog if you lack one.
Week 4: reviews, community, measurement
- Email recent customers asking for an honest G2 review, with no incentive attached.
- Turn the ten most common support questions into public docs pages.
- Run the question set again and compare with the baseline: who is named now, and which cited sources are you still missing from?
Then repeat the run monthly and add one comparison or use-case page a month. AI visibility for startups covers the stage before you have reviews; AI visibility for B2B services covers service firms rather than software.
How SearchCrow helps
SearchCrow runs the monthly measurement for you. The free check asks ChatGPT and Perplexity five questions a buyer might ask about your product and shows whether you were named and who was named instead. Paid plans track up to 150 questions nationally across ChatGPT, Gemini, Perplexity, Claude, Copilot, Google AI Overviews and Google AI Mode, show share of voice against up to 25 named competitors, list the sources each assistant cites that you are missing from, and email you when a competitor overtakes you. Everything in this guide works without it.
FAQ
How do I get my SaaS recommended by ChatGPT? Make your site eligible first: OpenAI says to allow its OAI-SearchBot crawler and make sure your host or CDN allows traffic from its published IP addresses, and that placement is not guaranteed. Then give it something to find: pricing, integrations and comparison pages in plain text, docs that need no login, and genuine reviews on sites such as G2. Ask ChatGPT your buyers' questions each month and watch what changes.
Do G2 and Capterra block AI crawlers? Mostly no, as of 10 November 2026. G2's robots.txt gives GPTBot, ClaudeBot, CCBot, Amazonbot, Meta-ExternalAgent and Applebot-Extended extra restrictions (individual review pages, three-way compare pages and non-English sections) but does not block them from the site; OAI-SearchBot, PerplexityBot and Claude-SearchBot are not named at all. Capterra's robots.txt names no OpenAI, Anthropic, Perplexity or Google token and fully blocks only a list of scrapers.
Should my SaaS pricing page show actual prices? If you want AI assistants to answer pricing questions accurately, yes, in plain text. Assistants answer by reading pages their crawlers can fetch, and Google says generative AI features use publicly accessible, crawlable content. A page that says only Contact sales gives an assistant nothing to quote, so it will quote a review site, a competitor's comparison page or an old forum thread instead.
Do I need llms.txt for my SaaS? It is optional. llms.txt is a proposal, first published in September 2024, for a Markdown file at /llms.txt that summarises a site for language models. Google says it does not need new machine-readable files or Markdown and that Google Search ignores them. G2 and Capterra both point to one from their robots.txt files, so it costs little to add after the pages in this guide exist, but it is not what gets you recommended.
Can documentation behind a login be used by AI assistants? Not by search crawlers. OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot fetch pages the way an anonymous visitor would, so anything behind a login or an email gate is invisible to them. Google also notes that not all bots can run JavaScript, so docs rendered entirely in the browser may be unreadable too. Put at least your integrations, API overview and getting-started docs on public, server-rendered pages.
Sources
- Google Search Central, Optimizing for generative AI features on Google Search
- Google Search Central, AI features and your website
- Google Search Central, Software app (SoftwareApplication) structured data
- Google Search Central, Understand the JavaScript SEO basics
- Google Search Console Help, Search generative AI control
- Google Search Central, Google's common crawlers (Google-Extended)
- OpenAI, Overview of OpenAI crawlers
- OpenAI Help Center, Searching the web with ChatGPT
- Perplexity, Perplexity crawlers
- Perplexity, Introducing the Perplexity Search API (25 September 2025)
- Anthropic, Does Anthropic crawl data from the web, and how can site owners block the crawler?
- Anthropic, Enabling and using web search
- G2, G2 Profiles
- G2, robots.txt
- Capterra, robots.txt
- llmstxt.org, The /llms.txt file