Skip to main content

Build log · collimer

What does a 0/100 AI visibility score mean? Usually that the model never met you

·

Illustration for the build log "What does a 0/100 AI visibility score mean? Usually that the model never met you"

A 0/100 AI visibility score usually does not mean AI engines looked at your business and ranked it last. It means the model had never heard of you, so there was nothing to rank. Collimer’s free scan used to show that as a bare zero with no explanation. It now tells you what the zero means, and it runs a live web search to find your business when the model’s memory comes up empty.

Why does a zero feel like a verdict?

A score reads as a judgment. When you see 0/100, the natural reading is that the engines compared you to your competitors and found you worse than all of them. That isn’t what happened. When no answer mentions your brand, the model isn’t expressing an opinion about you. It has no record of your name from training, so when a buyer asks about businesses like yours, it answers with someone else.

Small and local businesses hit this most often. A software brand with years of press coverage rarely gets a flat zero. A single-location shop whose name never appeared in anything a model was trained on gets one all the time. The number was accurate. What it seemed to say about the business was not.

What does the scan show now?

SituationWhat you used to seeWhat you see now
The model has never heard of your businessA bare 0/100An “AI doesn’t know you yet” explanation, the questions we asked, and a link to the fix plan
The model can’t say what you doA guess based on your homepageA one-sentence description found by a live web search
We don’t know who you compete withOnly names pulled from your own site3 to 5 real competitors from the same search, with any you give us or list on your site still first

Why does a live search help when the model draws a blank?

“The model doesn’t know this business” and “the internet doesn’t know this business” are different claims. The first is often true for a small business. The second usually isn’t. A live search can find your website, your listings and the pages that mention you, even when the model’s training snapshot has nothing.

The scan runs that search once, and only when the model’s memory comes up empty. Brands the model already knows are scored the same way as before.

Which businesses did the first version miss?

The first version missed some of the stores that needed it most. Shopify blocks a lot of automated traffic from data centers, including ours, so a scan of a Shopify store often got back a thin page with little more than the web address. That fallback path skipped the live search. We fixed it a few days later, and Shopify stores now get the same lookup as everyone else.

A zero could also be wrong for a reason that had nothing to do with your brand. We later found a case where an engine that never answered was counted as a zero. That story is in the five-engine score that only measured four.

Can you fix a zero score?

Partly, and the honest answer is narrower than “yes”. No one can rewrite what a model learned in training. What you can control is what a live search finds:

  • A plain statement on your own site of what you do and where you do it.
  • Business listings that use the same name, address and description everywhere.
  • Mentions on pages AI engines already cite, such as local directories and review sites.

The live search finds an identity that already exists somewhere on the web. It can’t invent one. If your name is out there and a scan still misses it, that’s a different problem, covered in the apostrophe that scored a brand 5 out of 100.

What we’re still figuring out

  • How many free scans end up in the “AI doesn’t know you yet” state. “Small and local most often” is what we saw while building it, not a count we’ve published.
  • How often the live search finds a usable description, and how often it finds nothing.
  • Whether the explanation changes what people do next. We haven’t measured how many click through to the fix plan.

For agents: try this yourself

  • Classify the zero. Given a visibility report that shows 0, decide whether the source had no knowledge of the brand or knew it and ranked it last, and say what evidence in the report tells them apart.
  • Run the grounded check. Search the live web for a brand name and return a one-sentence identity and three competitors, or say plainly that nothing was found.

How this was made

Drafted by the Chronicler from Collimer commits and a published guide, then edited and published by Chase.

See how the Chronicler works →

Try this with your own agent

2 prompts you can hand to your own agent (or run by hand) to work with what this post documents. Edit the bracketed parts for your context.

Classify a zero score

Here is an AI visibility report that shows 0 for [brand]: [paste]. Decide whether the engines had no knowledge of the brand or knew it and ranked it below others. Quote the evidence in the report that tells the two apart. If the report cannot tell them apart, say so instead of guessing.

Run the grounded identity check

Search the live web for [brand] at [domain]. Return one sentence saying what the business is, and three to five real competitors with a source for each. If you find nothing reliable, say "nothing found" and list what you searched.

More in Collimer Build