Study: 82% of Local Businesses Are Invisible to AI Recommendations

Key Findings

• 82% of businesses (243 out of 298) never appeared in any AI recommendation across three models and three prompts each.
• Only 4 businesses out of 298 were recognized by all three AI platforms. Universal AI visibility is near-impossible.
• AI models disagreed 100% of the time on which businesses to recommend. The AI platform a consumer uses determines which businesses they find.
• Claude recognized 3x more businesses than Gemini (37 vs. 12). The gap between models is dramatic.
• Gemini was completely blind to two entire industries (HVAC and plastic surgery). Zero mentions across 98 businesses.
• The highest-reviewed business in every industry scored zero. Google reviews have no detectable impact on AI recommendations.

Dataset Summary

Industry Breakdown

AI invisibility was consistent across all six industries, but the severity varied. Roofers were the most invisible at 94%, while plastic surgeons fared best at 71%.

The 23-point spread (71% to 94%) suggests that some industries have stronger representation in AI training data than others. But even the best-performing industry still has nearly three out of four businesses missing from AI responses.

How AI Models Differ

The three AI platforms produced dramatically different results. Claude recognized the most businesses, ChatGPT fell in the middle, and Gemini trailed far behind.

Gemini was unable to identify a single business from our sample in two entire industries: HVAC companies and plastic surgeons. That means 98 businesses were scanned across those two industries, and Gemini mentioned none of them. A consumer using Gemini for recommendations in those categories would receive a completely different set of suggestions than someone using ChatGPT or Claude.

Perhaps the most striking finding across the entire study is the disagreement rate: 100%. In every single industry, the three models recommended different sets of businesses. Not one industry produced a consensus. This is not a case of slight variation. The models are working from fundamentally different understandings of local markets.

What this means in practice: the AI assistant a consumer happens to use determines which businesses they discover. A homeowner asking ChatGPT for a roofer will see different companies than one asking Claude. Neither will see the list that Gemini generates. There is no single "AI result." There are three separate realities, and they barely overlap.

Appearance Distribution

Of the 298 businesses in our sample, here is how many appeared in zero, one, two, or all three AI models:

The Reviews Disconnect

One of the most consistent findings across all six industries: Google review counts have no detectable relationship with AI visibility. The highest-reviewed business in every single industry scored zero.

Combined, the top-reviewed businesses in these three industries have 8,022 Google reviews and an AI visibility score of zero. Meanwhile, lesser-reviewed competitors hold the top visibility spots.

This is not a statistical coincidence. It repeated in every single industry we tested. Traditional online reputation signals like Google reviews, star ratings, and review volume do not appear in the data AI models use to generate recommendations.

What Surprised Us

We expected variation between models. We did not expect entire industries to be erased.

Gemini, built by Google, could not identify a single HVAC company or plastic surgeon in Austin. These are not obscure businesses. They have Google Business profiles, hundreds of Google reviews, and active websites indexed by Google Search. Yet Google's own AI model returned zero mentions for 98 businesses across those two industries.

This suggests that the data powering Google Search and the data powering Google's AI recommendations operate independently. A strong Google Search presence does not carry over into Gemini's awareness. For businesses that have invested years optimizing for Google, this is a significant blind spot.

We were also surprised by the consistency of the reviews disconnect. In a single industry, it might seem like a coincidence. But when the highest-reviewed business scores zero in every industry you test, the pattern becomes undeniable. Whatever signals AI models use to generate recommendations, Google review volume is not one of them.

Why This Matters

AI Compresses Discovery

When a consumer searches Google for "best dentist in Austin," they see a map pack, organic results, ads, and review sites. Dozens of businesses can appear across the first few pages. The consumer browses, compares, and filters.

When that same consumer asks ChatGPT the same question, they get a single list of three to five names. No map pack. No second page. No sponsored results. AI compresses local discovery into a handful of recommendations, and everything else disappears.

This is a structural shift in how consumers find local businesses. Traditional search distributes attention across many results. AI concentrates it on a few. In our study, each model typically named fewer than 10 businesses per industry. Out of 50 competitors, AI surfaced roughly 5. The other 45 did not exist in the response.

A New Competitive Dimension

For local businesses, this compression creates a new competitive reality. Visibility in traditional search results does not guarantee visibility in AI recommendations. A business that invested years building Google reviews, earning backlinks, and climbing organic rankings may find that none of that effort translates to the AI-generated shortlist.

The businesses that do appear in AI responses gain a disproportionate advantage. When AI names five companies and excludes forty-five, the recommended businesses capture attention without competing against ads, review sites, or aggregators. The shortlist is the entire result.

Three Implications

• The platform a consumer uses matters. A patient asking Claude for dentist recommendations will see a completely different list than one asking ChatGPT. Businesses visible on one platform may be invisible on another.
• AI visibility is becoming a new competitive factor. As AI usage grows, being on the shortlist provides an advantage that traditional marketing cannot replicate. The businesses AI recommends will receive an increasing share of consumer attention.
• Most businesses do not know their AI visibility status. Unlike Google rankings, which can be checked with a simple search, AI visibility requires structured testing across multiple models. The gap between perceived and actual visibility is likely enormous.

This is not a temporary glitch. As AI assistants continue to gain users, the compression of local discovery will only intensify. The question for every local business is not whether AI will affect their visibility. It is whether they are on the shortlist or excluded from it.

Methodology

This study was conducted in March 2026 using the following procedure:

• Business sourcing: Businesses were identified via the Google Places API using industry-specific search terms in Austin, TX. Results were filtered for active websites and deduplicated by domain.
• Sample selection: The top businesses by Google review count in each industry were selected, yielding 298 businesses across 6 industries.
• AI scanning: Three natural-language prompts were sent to each AI model (ChatGPT, Gemini, Claude) for each business. Prompts were written to mirror how real consumers ask AI for local service recommendations. No prompt engineering or specialty modifiers were used.
• Response analysis: AI responses were analyzed for specific business mentions, recommendation context, and accuracy. Each business received a visibility score (0-100) based on mention frequency, recommendation quality, and cross-model consistency.
• Total queries: 2,682 AI queries were executed across all businesses, models, and prompts.

Individual Industry Reports

Detailed findings for each industry are available in the individual reports:

Check Your Visibility

This report provides a market-level view of AI visibility for local businesses. Your specific situation may differ.

• You could be one of the 18% that AI already recognizes
• Competitors may be visible on platforms you have not checked
• AI may describe your services inaccurately or reference outdated information

Citation

If you reference this study in an article, report, or presentation, please link to the original report: