What the GEO checker actually checks, and why

The GEO checker does two things, and it's worth being precise about both. First, it looks at the page itself for the signals that make content easy for any machine to parse: a real title, a meta description, structured data, a sane heading structure, and enough actual text to work with. None of that is new — it's closer to classic SEO hygiene than anything AI-specific.

Second, and this is the part that's actually new, it hands the page's text to Claude with no other context and asks it to do exactly what an AI answer engine would do if someone asked it for a recommendation: read the page once, and decide what it just learned. Can it state what you do? Who you help? Did it find a specific, concrete result, or just a claim? Would it actually put your name forward? That's not a heuristic — it's the real behavior we're trying to score, run directly.

An earlier, more literal version of this idea would be to ask a handful of AI assistants live, "who's a good freelance [your field]" and see if you come up. We didn't build that, on purpose. For the overwhelming majority of freelancers, the honest answer to that question is no, regardless of how good their portfolio is — general-purpose assistants haven't memorized most individual freelancers, and a tool that just told you that over and over wouldn't be useful, it would just be discouraging.

The question that's actually actionable is different: if an AI reads your page right now, does the page itself give it enough to work with? That's the thing you can fix this afternoon, and it's the thing that matters once these assistants do start pulling from the web in real time to answer exactly this kind of question — which is already happening for some of them.

It's deliberately simple: nine checks, each pass or fail, score is the percentage you pass. No hidden weighting, no checks you can't see. Five are on-page signals, four come from Claude's read of the page. We'd rather the checker be legible and a little blunt than dress up a vague feeling as a precise-looking number.

It's also brand new. If a score or a fix suggestion looks wrong for your specific page, that's useful to know — reply to the report email or reach out, and it'll get sharper.