When AI systems confused NuGEO with other companies

Could selected AI services identify NuGEO as the company at nugeo.com, describe its work accurately, and tell it apart from similarly named organizations?

We ran the audit before promoting NuGEO more widely. A mention does no good if the answer points to the wrong company, category, or domain.

On 29 August 2026, NuGEO ran the same 30 questions across five selected AI systems. They covered category discovery, comparisons, recommendations, reviews, alternatives, brand understanding, competitive positioning, and poor-fit cases.

Four systems completed the set. One returned no completed answers, so we excluded it from the outcome rates instead of scoring it as a negative result. The crawl inspected nine canonical public pages and one internally linked homepage state.

Many answers used the NuGEO name for another company. We saw nutrition products, a similarly named GEO platform, fleet software on another country-code domain, geotechnical organizations, review software, and a domain listing.

That made the headline mention rate misleading. We now score whether an answer has the right company and category, not just the right spelling.

• The site described the report but said little about the team, track record, or selective client model behind it.
• The site had no machine-readable Organization or WebSite identity.
• The public method was inspectable, but there was no worked NuGEO example connecting a saved question to a finding and acceptance test.
• The Research hub contained one guide and little dated documentation of the work itself.

We now repeat the same core facts across the public site: NuGEO is a boutique, invite-only GEO intelligence and implementation service; its service has supported companies for more than two years; and the delivery team is distributed across the United States and Canada.

The homepage includes Organization and WebSite markup. Foundation is identified as a service, and each research article shows its author and date. None of the markup says more than the page itself.

1. Preserve the question set.Use the same 30 questions and the same selected systems.
2. Wait for discovery.Confirm the changed pages are crawlable and inspected before interpreting a rerun.
3. Score the correct entity.Separate a literal name mention from an accurate company, domain, category, and operator match.
4. Inspect citations.Record whether an answer cites the canonical NuGEO page that supports its description.
5. Keep unavailable systems out of rates.Show provider failure as missing coverage, not poor brand performance.

This is one sample from one day. It does not cover private conversations, every model, every location, or every source. Answers can change even when the site does not, and a site change does not guarantee a mention or citation.

The baseline gives us something narrower and useful: a record of the test, the change, and the result of the same test later.