Free GEO tools, sorted by what they can actually see

One job is access: can a crawler fetch the site, and is the brand written the same way in the title and the H1? Another job is sampling: what does a specific answer product say when you ask a specific prompt today? A third job is memory: did that answer change since last week? Free tools are good at the first, uneven at the second, and usually honest when they refuse the third.

Benchline is built for the first job, plus a notebook for the second. You get a permanent /r/… report, a horizontal prompt pack, and a paste scorer. You do not get a historical chart.

Use these in this order

1. Run the citability check. Fix blocked crawlers, a missing llms.txt, and a title that never says the brand. The point values are on the methodology page.
2. Validate JSON-LD in a schema validator if the report says Organization or FAQ markup is missing and you thought it was there. Our parser reads JSON-LD and microdata on the homepage only.
3. Run four prompts yourself: one category prompt, one alternatives prompt, a “what is this brand” prompt, and an “is it worth it” prompt. Paste each reply. This walkthrough is the whole method.
4. If you need ongoing multi-engine share of voice, look at a paid suite that publishes that product. Ahrefs and Semrush both offer AI visibility checks with that shape. Copying it on every free hit is out of scope here.

What we left out on purpose

No invented comparison scores, no traffic estimates, and no “AI search volume”. If a number is not computed from your crawl or from text you pasted, it is not on the page. The longer definition of the work is what generative engine optimization is.