In short
- FoundByChat's October 2026 study (4,203 search results in 211 ChatGPT answers) found ChatGPT cited retrieved vendor pricing pages 51.2% of the time, product pages 28.4%, and docs 23.0%.
- Company pages were cited 20.1% of the time when retrieved, versus 3.9% for third-party pages about the company and 7.5% for listicles and reviews.
- Vercel and MERJ (December 2024) found major AI crawlers like OpenAI's OAI-SearchBot do not render JavaScript, so content that requires JavaScript to render may not be present for OAI-SearchBot.
- Our October 2026 study of 218 answers showed position 1 in ChatGPT's first search was cited 55.5% of the time, versus 0.4% at position 16+.
Why SaaS SEO shifts from listicles to your own pricing, product, and docs
In FoundByChat's study, pages on a company's own site were cited more often when retrieved than third-party pages about the company. In FoundByChat's October 2026 study of 4,203 labelled search results across 211 ChatGPT answers, pages on a company's own website were cited 20.1% of the time when retrieved, against 3.9% for third-party pages about that company (an odds ratio of 13.1). To understand what AI SEO is for a software business, look at which page types get quoted when retrieved.
| Page type | Share cited when retrieved | Source |
|---|---|---|
| Vendor pricing pages | 51.2% | FoundByChat study (Oct 2026) |
| Vendor product pages | 28.4% | FoundByChat study (Oct 2026) |
| Docs and help pages | 23.0% | FoundByChat study (Oct 2026) |
| Vendor blog posts | 14.5% | FoundByChat study (Oct 2026) |
| Third-party listicles and reviews | 7.5% | FoundByChat study (Oct 2026) |
| "A vs B" comparison pages | 4.2% | FoundByChat study (Oct 2026) |
| PDFs | 0.4% | FoundByChat study (Oct 2026) |
In a separate October 2026 FoundByChat study of 218 ChatGPT answers with web search, ChatGPT cited 2.2 different URLs per answer on average and cited 10.3% of retrieved web results. Reddit made up 28% of retrieved results, but accounted for only 2 of the 2,462 citations.
Step 1: Put plain facts on your pricing, product, and comparison pages
Consider putting concrete numbers and specifications in the opening text of your pricing and product pages.
- Publish numbers on your pricing page: State plan prices, user minimums, and usage limits in plain HTML text rather than hiding them behind interactive calculators.
- State facts up front on product pages: Put what the product does and its specs in the opening sentence and check whether menu or cookie text leaks into the ~200-character snippet that ChatGPT's search keeps.
- Add quotations, statistics, and sources to blog and comparison posts: The November 2023 Generative Engine Optimization (GEO) paper by Pranjal Aggarwal and colleagues at IIT Delhi and Princeton (KDD 2024) tested 10,000 queries on GEO-bench and found that adding quotations gave a +41% gain on its position-adjusted word count metric (27.2 versus a 19.3 baseline), adding statistics scored 25.2, and citing sources scored 24.6, while keyword stuffing scored below baseline at 17.7. Lower-ranked pages gained most: rank 5 pages gained +115.1% from citing sources, while rank 1 pages saw -30.3%.
- Move specs out of PDFs: Because PDFs were cited 0.4% of the time when retrieved in our October 2026 study, publish security sheets and API limits as HTML pages.
Step 2: Serve docs and marketing pages as server-rendered HTML for OAI-SearchBot
Allow OAI-SearchBot if you want your site to appear in ChatGPT search answers, and consider serving important text in HTML without requiring JavaScript.
In December 2024, Vercel and MERJ analysed AI crawler traffic and found that "none of the major AI crawlers currently render JavaScript", naming OpenAI (OAI-SearchBot, ChatGPT-User, GPTBot), Anthropic (ClaudeBot), Meta, ByteDance, and Perplexity. They download JavaScript files without running them (11.50% of ChatGPT's requests and 23.84% of Claude's were for JS files), while Google's Gemini uses Googlebot (which renders JavaScript) and Applebot renders through a browser.
OpenAI documents its bots at developers.openai.com/api/docs/bots (retrieved October 2026), and their settings are independent:
| Bot name | User agent string | Role | Effect of blocking in robots.txt |
|---|---|---|---|
| OAI-SearchBot | OAI-SearchBot/1.4; +https://openai.com/searchbot | Surfaces websites in ChatGPT search results | Site will not be shown in ChatGPT search answers (can still appear as navigational links) |
| GPTBot | GPTBot/1.4; +https://openai.com/gptbot | Crawls content that may be used in training generative AI foundation models | Tells OpenAI content should not be used for model training; does not remove site from search |
| ChatGPT-User | ChatGPT-User/1.0; +https://openai.com/bot | Handles certain user actions in ChatGPT and Custom GPTs | Not used to decide Search appearance; OpenAI says robots.txt rules may not apply to user-initiated actions |
| OAI-AdsBot | OAI-AdsBot/1.0 | Checks safety of pages submitted as ChatGPT ads (not used to train models) | Advertisers must allow it; OAI-SearchBot is also recommended |
After a robots.txt change, it can take about 24 hours for ChatGPT search to adjust. OpenAI's publisher FAQ notes a URL blocked in robots.txt but learned from a third-party search provider may still show as a link and page title; a readable noindex tag prevents that.
Step 3: Rank high in ChatGPT's first search and track buyer questions
Citation rates drop sharply after the top search results, which is why SEO in 2026 combines search ranking with direct answer structure. In FoundByChat's October 2026 study of 218 ChatGPT answers with web search, the citation rate by position in ChatGPT's first search was:
- Position 1: 55.5% (95% CI 49.1 to 61.9%)
- Position 2: 34.9%
- Position 3: 27.5%
- Positions 4 and 5: 14.4%
- Positions 6 to 10: 8.1%
- Positions 11 to 15: 2.4%
- Position 16 and lower: 0.4%
To inspect your key pages, run your URLs through the free GEO audit at foundbychat.com/geo-audit (no account needed, 5 audits a day). It fetches your page with the OAI-SearchBot user agent, rebuilds the title and ~200-character snippet ChatGPT's search keeps, looks the page up in ChatGPT, and returns a score out of 100 with a letter grade covering bot protection, HTML text without JavaScript, menu or cookie text leaking into the snippet, JSON-LD, llms.txt, robots.txt rules, and noindex.
To monitor buyer queries over time, add your website and customer questions to FoundByChat (free to start). It asks ChatGPT those questions with web search on every 2 hours from the US (English) or the Netherlands (Dutch), records whether your pages are "Found" or "Cited" and their search position, lists the title and snippet kept for your "Pages in ChatGPT's index", and emails you when something changes. Track referrals via utm_source=chatgpt.com.
FOUNDBYCHAT
Know the moment ChatGPT picks up your pages
We ask ChatGPT your questions every 2 hours and mail you what its index picked up or rewrote. Free to start: sign in with Google or an email link.
Questions
Does blocking GPTBot hurt our SaaS website in ChatGPT search?
No. Blocking GPTBot in robots.txt only tells OpenAI not to use your content to train generative AI foundation models. OpenAI's crawler documentation (October 2026) states that bot settings are independent, and OAI-SearchBot surfaces websites in ChatGPT search. Keep OAI-SearchBot allowed in robots.txt and bot protection settings.
What did the study find about pricing pages and blog posts?
In FoundByChat's October 2026 study of 4,203 retrieved results in 211 answers, vendor pricing pages were cited 51.2% of the time when retrieved, versus 14.5% for vendor blog posts. Consider putting exact prices and limits in plain HTML.
Can ChatGPT read our React or single-page documentation site?
OpenAI publishes no statement on JavaScript execution, but a December 2024 Vercel and MERJ study found that none of the major AI crawlers, including OAI-SearchBot, ChatGPT-User, and GPTBot, render JavaScript. They download .js files (11.50% of ChatGPT's requests) without executing them. Test if your pages have readable text without JavaScript using the free GEO audit.
Should we still build "Alternative to X" and "A vs B" comparison pages?
In FoundByChat's October 2026 study of 4,203 retrieved results, "A vs B" comparison pages were cited 4.2% of the time when retrieved, versus 28.4% for vendor product pages and 23.0% for docs. In the study, product, pricing, and docs pages had higher citation rates when retrieved than comparison pages.
Sources
- OpenAI: Overview of OpenAI Crawlers (OAI-SearchBot, GPTBot, ChatGPT-User)
- OpenAI Help Center: Publishers and Developers FAQ
- OpenAI Help Center: Advertiser Guidance for Allowing OpenAI Web Crawlers
- Search Engine Journal: OpenAI Says Robots.txt May Not Apply To ChatGPT's Fetch Bot
- Vercel: The rise of the AI crawler (Vercel and MERJ study on JavaScript rendering)
- arXiv (KDD 2024): GEO: Generative Engine Optimization by Aggarwal et al.