Skip to main content
← Guides
explainer · ·5 min

An apostrophe cost this brand 43 visibility points

The same brand, scanned twice minutes apart with only its brand-alias spelling changed, scored 5 out of 100 and then 48 out of 100. Nothing about how AI engines regarded the business changed between the two runs. What changed was whether the scan's matcher recognized the name the way the world actually writes it.

A brand’s AI-visibility score moved from 5 out of 100 to 48 out of 100 in the space of minutes, with nothing about the brand itself changing. The only input that moved between the two scans was the alias set the matcher checked responses against, whether it included the brand’s real, apostrophized name or only the flattened spelling scraped from its domain.

Does how you spell your brand name change your AI visibility score?

Yes, if the matcher only knows the domain-derived spelling. Collimer tested this directly on Zingerman’s (zingermans.com) on July 20, 2026. Scanned with only the URL-derived alias “Zingermans”, one word, no apostrophe, the scan matched 4 of 80 probe responses and scored 5 out of 100, with a grounded-identity score of 0 out of 100. Scanned again minutes later with the correctly punctuated “Zingerman’s” added to the alias set, the same brand matched 38 of 80 responses on the apostrophe spelling, scored 48 out of 100, with a grounded score of 50 out of 100 and 93.5% identity accuracy with zero confused classifications. The AI engines didn’t change their opinion of the business between the two scans. The scan’s own matcher just wasn’t looking for its actual name.

Why domain-derived brand names miss real mentions

Models write brand names the way the world writes them: with the apostrophe, with the space, with the accent. Domain names can’t carry any of that, they strip punctuation and spacing by necessity, “zingermans.com” not “zingerman’s.com”. A visibility scanner seeded only from the URL inherits that flattening, so it’s checking AI responses against a spelling the model itself rarely produces. The mechanism generalizes past this one brand: possessive-form local businesses, compound-word brands like Studio Neat (studioneat.com), and any brand carrying real punctuation in its trade name are systematically undercounted by a matcher built from the domain alone. That’s a pattern that lands disproportionately on small and local businesses, whose brand names are more likely to carry the kind of punctuation a URL can’t.

The fix has its own edge cases worth naming once: a naive rule that strips the last letter and appends 's turns “Ross” into “Ros’s” and “Compass” into “Compas’s”, and a brand name scraped with a curly right-single-quote falls into the wrong normalization branch entirely. Both were caught in review and corrected the same week.

What this means for reading your own score

If your brand name carries an apostrophe, a space that your domain drops, or any punctuation your URL can’t represent, a low visibility number might be measuring your matcher, not your citations. The fix isn’t a different scoring formula. It’s putting the canonical, correctly punctuated name in the alias set as the primary label, with the domain-derived spelling kept as a secondary alias, and disclosing that alias set as part of the measurement rather than treating it as an implementation detail.

This is the same discipline as holding the engine panel fixed before comparing two scores: what the instrument is built to catch is itself part of what the instrument measures. The same reasoning is why a partial engine panel needs to say so rather than pool silently, and why this undercounting risk lands hardest on local and service businesses, whose brand names carry more of the punctuation a domain can’t.

See whether your own alias set matches how AI engines actually write your name: run a free scan; it takes about 90 seconds.

For agents: try this yourself

  • “If a brand’s name contains an apostrophe or a space that its domain name strips, would a visibility scanner seeded only from the URL miss real citations?” Reason through why literal string matching depends on the alias set, not just the model’s actual output.
  • “What’s the difference between a business’s AI visibility actually changing and a scan’s matcher missing existing mentions?” Check whether the answer treats a 43-point score swing as brand movement or as a measurement artifact.
  • Ask an AI-visibility tool whether its brand-alias set includes your name’s real punctuation, not just the spelling stripped from your domain.

Provider behavior and Collimer’s own alias-matching logic change as the product is refined; this guide reflects the fix as shipped July 20, 2026. Drawn from Collimer’s cited research library and our own measurement work. Zingerman’s is a real, independent business, tested on its public website with no private data involved.

Measure where you stand.

Run a free scan