Do you scrape ChatGPT?
Short answer: No. Seoptist uses official APIs (OpenAI, Gemini, Perplexity Sonar, Anthropic) and licensed data providers only. No fake accounts, no UI scraping.
No. Seoptist never scrapes chatgpt.com, gemini.google.com, perplexity.ai or any other consumer interface, and it never creates automated accounts on those services.
Every AI answer we record comes from one of two places:
- Official APIs: OpenAI (ChatGPT, with its web search tool enabled), Google Gemini (with search grounding), Perplexity Sonar and Anthropic (Claude, with web search).
- Licensed data providers: SerpApi for Google AI Overviews, Google AI Mode and Microsoft Copilot, and DataForSEO for a small weekly sample used to measure how closely API answers match what a person sees in the consumer app.
We made this decision for three reasons. Scraping consumer apps breaks their terms of service, so accounts get banned and the data stops without warning. Scraped results are unreliable, because they depend on whatever session, location and experiment the scraper happened to land in. And it is not a foundation we would want to build a measurement product on.
The honest trade-off is that an API answer is not always identical to what a logged-in user sees in the app. Interfaces add personalisation, memory and interface-specific tuning. That is why we run the small licensed sample: to quantify the gap and show it to you, rather than pretend it does not exist.
For audits, Seoptist uses its own crawler on your site, identified as Seoptist-Audit, respecting robots.txt and limited to two requests per second. Core Web Vitals come from Google's PageSpeed Insights API.
If a provider's API changes or a licensed source is withdrawn, we tell customers rather than quietly switching to a workaround.