# Get found in ChatGPT, Claude and Perplexity | Propel Source: https://runpropel.com/get-found Get found # Be the answer when people ask AI. Your buyers ask AI assistants which tool to use. Propel checks what ChatGPT, Claude, Gemini, Perplexity, Grok and Bing say about you every week, fixes what is missing, and shows you whether each fix worked, with a range rather than a guess. [Join the waitlist →](https://runpropel.com/waitlist)[See pricing](https://runpropel.com/pricing) Measured properly ## What AI says about you, with a range. It asks the questions your buyers ask, many times over, across ChatGPT, Claude, Gemini, Perplexity, Grok and Bing . Each result shows how often you're mentioned, the range, and how many answers it's based on. A single screenshot proves nothing. A range does. Results · Get found Example 31% of ChatGPT answers mention you range 24–38% · 120 answers · read Monday Last month 18% (12–25%) Change Up, outside the range of luck Coding assistants ## What the coding assistants pick. Later Not in the public beta. This switches on after it. If you make something developers use, the decision often happens inside the editor. Propel runs Claude Code, Codex, Cursor and GitHub Copilot on real tasks in your category and records which tools they reach for. Coding lens · add currency conversion Example Claude Code picked another library Codex picked Ratewise Cursor picked another library GitHub Copilot picked another library Example task, run in a sandbox True or not ## It checks what they get wrong. Old pricing, a missing feature, a competitor's limit pinned on you. Propel checks answers against what's true about your product and shows you each mistake, with the source that proves it. Fact check Example Gemini says "Ratewise has no free plan." What's true A free plan with 1,000 requests a day. ratewise.io/pricing Who gets quoted ## Which pages do AI answers quote, and why not yours? Each answer comes with the pages the assistant leaned on. Propel sorts them into your pages, your competitors' and everyone else's, and sees which lists and reviews name your rivals but not you. Then it ranks the week's fixes: how many answers are at stake, how sure it is, and how much work each fix takes. The top of the list is where to start. Quoted this week · 120 answers Example A "best testing tools" list names two rivals, not you Fix #1 · pitch drafted A review site's comparison quotes your old pricing Fix #2 · correction drafted Your docs: why specs your own page Quoted in 9 of 120 Then it fixes ## It fixes what's missing, then proves the fix worked. It writes the page the answers are missing. A ready-made connector, so ChatGPT and Claude can use your product directly, is coming. It writes to the sites that get quoted, and fills in listings in the directories that matter. Each goes out after your yes; on Product Hunt, you press submit. Then it compares the questions it worked on with similar ones it didn't touch, so a change is shown as caused by the fix, not by luck. Fixes this month Example The page answers were missing Published Correction to a page AI quotes Waiting for your yes Did the page work? Yes: +14 points, outside luck And classic search ## Articles on your own site, measured. Articles and comparison pages, published straight to your site and measured in Google Search Console. Which AI crawlers read your site, from your hosting logs, comes later. Publishes to WordPress Webflow Ghost Shopify Wix Contentful Sanity Strapi Notion Hashnode dev.to any webhook ## No tricks sold as results. Some things get sold as "AI visibility" with no evidence they work. Propel will set them up if you ask, labelled "no measured effect". It won't promise what it hasn't measured. Want it done this month? [Propel Growth Desk runs it for you](https://runpropel.com/growth-desk), and a person approves every piece. ## Which AI assistants does Propel check, and how? ### Is this what people call AEO or GEO? Yes. Answer engine optimisation and generative engine optimisation are two names for the same goal: being the answer when someone asks an AI assistant. Propel measures it properly, fixes what is missing, and shows whether each fix worked. ### Which AI assistants does Propel check? ChatGPT, Claude, Gemini, Perplexity, Grok and Bing, through their official interfaces with web search on. Running Claude Code, Codex, Cursor and GitHub Copilot on real tasks, to see which tools they pick, comes later. ### Does it check Google's AI Overviews? Not as a rate yet, and Propel says so. Google's AI Overviews, AI Mode and the ChatGPT app aren't counted in how often you are named. Where the check is set up, a sample of the ChatGPT app's answers, and of Google's, is read only to see how far each app agrees with its official interface. Gemini is measured through its own interface. ### Why ranges instead of a single number? AI answers change every time you ask. One screenshot proves nothing. Propel asks many times and shows how often you are mentioned with a range and the number of answers behind it, so you can tell a real change from luck. ### Do llms.txt files and schema help? There is no good evidence yet that they change AI answers. Propel will set them up if you ask, labelled 'no measured effect'. It won't sell you what it hasn't measured.