You don't need a paid tool to know if AI recommends you. Build a 20-prompt buyer panel, run it in ChatGPT, Claude, and Perplexity, track 3 metrics monthly: visibility rate, position, citation share. 30 minutes a month. And when the verdict is "invisible": the cited sources are mostly Reddit threads, and that's fixable.
The 4-Step Method
Write 20-50 prompts your buyers actually type: "best [category] for [use case]" (5-10 variants), "[competitor] alternatives" (one per competitor), "[category] under $X", "is [category] worth it for [audience]", and 2-3 problem-phrased prompts ("how do I stop doing X manually"). Save them in a spreadsheet, one row each.
Run every prompt in ChatGPT, Claude, and Perplexity, each in a fresh session (no memory contamination). Columns per engine: mentioned yes/no, position (1st recommendation, listed, absent), and sources cited when visible.
Visibility rate = prompts mentioning you / total prompts. Average position = are you the answer or the afterthought. Citation share = which domains the AI cites; count how many are Reddit threads (usually the plurality) and whether you appear in any of them.
Same panel, same engines, first week of each month. AI answers vary run to run, so trust the trend across 20+ prompts, never a single answer. A visibility rate moving 10% to 30% over a quarter is a real signal.
Reading the Results
Whatever the score, the fix path runs through the same place: the sources column. For commercial prompts, those sources are dominated by Reddit (~40% of LLM citations). Reppit AI ($25/mo) finds the threads in your niche worth joining and drafts the replies; you post manually. Check monthly, act daily. The strategy layer: how to get mentioned by ChatGPT.
If you outgrow the spreadsheet (agencies, many brands, daily tracking), that's when paid measurement tools earn their price: the best GEO tools compared.
Frequently Asked Questions
How often should I check AI visibility?
Monthly is enough for a manual panel: citations move on a weeks-to-months cycle. Daily tracking only matters at agency/enterprise scale.
Should I test with or without web browsing enabled?
Both if you can: browsing-on reflects live retrieval (your recent Reddit work shows up faster), browsing-off reflects training-data presence. Note which mode per column.
Is a free AI visibility checker tool accurate?
One-shot checkers give a snapshot from a handful of prompts, useful as a teaser, too noisy for decisions. A 20+ prompt panel you control beats any single-run report.
Reppit AI gets you into the Reddit threads AI engines cite. $25/mo.
Fix My Visibility