How to Read Your AI Recommendation Score (ARI)

What the score is actually measuring

Your AI Recommendation Score, or ARI, is a snapshot of whether AI models mention your business when someone asks a recommendation-style question. Something like "best plumber in [your town]" or "who should I call for emergency AC repair near me."

It is not a ranking against every competitor in the world, and it is not a guess at your reputation. It is a direct check: we ask the question, we see what ChatGPT, Gemini, Claude, and Perplexity actually say back, and we record whether you show up, how you're described, and how consistent that is across models.

Think of it less like a credit score and more like a report card on a specific set of questions. It tells you what happened when we asked, not a permanent fact about your business.

Why the number moves

AI models don't all pull from the same sources, and they don't all answer the same way twice. That's normal. One model might mention you clearly, another might not mention you at all, and a third might name you but get a detail wrong.

A few things genuinely affect whether a model has anything to say about you:

• •Whether your website has clear, structured information about what you do and where you do it
• •Whether basic technical signals (like schema markup and a clear meta description) make it easy for a page to be understood
• •Whether your business is described consistently across the pages that exist about you
• •Whether there's enough public information anywhere for a model to have formed an "opinion" in the first place

Your ARI can shift from one check to the next even if you've changed nothing, because the models themselves get updated or behave slightly differently run to run. That's expected. What matters more is the pattern over time, not any single check.

Reading the parts, not just the total

The overall ARI number is useful as a headline, but the more useful part of the report is underneath it: which models mentioned you, which didn't, and what they said when they did.

If you're mentioned by some models but not others, that's a signal your basic information exists but might not be structured or complete enough to be picked up everywhere. If you're not mentioned by any model, that usually points to a more fundamental gap, like thin website content or missing basic signals, rather than something you can fix with a quick tweak.

It's also worth reading what the models say about you, not just whether they say your name. A mention that gets your service area or specialty wrong is a different problem than a mention that's accurate but brief. Belcher shows you both, because "were we mentioned" and "were we described correctly" are genuinely different questions.

What Belcher checks, and what it doesn't

To be clear about how the score is built: Belcher asks AI models recommendation-style questions and records what comes back. It also reads your website directly to check things like schema markup, your meta description, heading structure, and page speed, since these are the kind of basic signals that make a page easier for a model to understand and use.

What Belcher doesn't do is scrape or monitor Reddit, social media, or review sites. If your visibility problem lives mainly in a place like that, our score won't capture it, and we'd rather tell you that plainly than imply otherwise.

What to actually do with your score

The point of ARI isn't to stare at a number, it's to find the gap between what's true about your business and what the models can currently see. Usually that gap is a website issue: missing schema, a vague meta description, or pages that never clearly state what you do and where.

Belcher can generate fix content, schema, FAQ blocks, meta snippets, that you paste into your own site to close some of those gaps. We don't publish anything for you or promise a guaranteed jump in the score afterward. What we can say is that clearer, more complete, more consistent information about your business is the thing most likely to help, and the score is there so you can check whether it did.