The Match Protocol™ Benchmark 2026 found that 80 of 100 U.S. small businesses were Ghosted by AI search, with an average Match Protocol Score™ of 50.9 out of 100. AI could read most of these businesses and verify them. It just rarely recommended them.
What did the Match Protocol™ Benchmark 2026 measure?
The benchmark measured how often ChatGPT, Google AI Overviews, Perplexity and Gemini recommend a small business when a buyer asks a question that never names it.
Modern Asset Lab scored 100 U.S. small businesses across 10 industries: HVAC, plumbing, roofing, dental, med spa, personal injury and family law, auto repair, independent restaurants, CPA and accounting, and landscaping. They sat in 11 metro areas across the South, West and Midwest. This is the first edition of the benchmark, run in October 2026.
Each business was run through the Match Protocol Score™ methodology: eight buyer questions, each asked of all four engines, for up to 32 blind buyer conversations per business. An average of 27.8 were answered.
What were the headline results?
The headline is that AI search leaves most small businesses out of the answer.
- Average score: 50.9 out of 100. Median: 49.5.
- 80 Ghosted (0 to 59), 13 Friend Zone (60 to 69), 5 First Date (70 to 79), 2 Going Steady (80 to 89), 0 The One (90 to 100).
- 36 were never recommended. 31 were never even named.
- 62 of 100 were recommended 3 times or fewer across an average of 27.8 conversations.
- Only 8 businesses were recommended in more than half of their conversations.
Nobody reached the top tier. The highest score in the benchmark was 88.
Why can AI read these businesses and still not recommend them?
The stage averages show the break: AI finds and trusts these businesses, then fails to match and recommend them.
Found averaged 80. Trusted averaged 88. Matched averaged 42, and Recommended averaged 18. The Recommended stage carries 35% of the score, so a weak result there drags the whole number down. A business can be easy to see and easy to verify and still be left out of the answer.
Recommendation is a choice. The engine has to connect your business to the buyer's question and find an answer it can lift. That is where most businesses fall apart.
Did any industry or city escape?
No. Among the industries large enough to quote, every one averaged under 60.
Plumbing scored highest at 56.6 and HVAC next at 54.9. Dental practices scored under 60 in 10 of 11 cases, auto repair in 9 of 10, roofing in 9 of 11 and med spas in 8 of 10. Among metro areas with 10 businesses, St. Louis led at 55.6 and Boise trailed at 44.9. Across the three Census regions, the Midwest averaged 54.3, the South 51.8 and the West 46.5. Being in a strong category or a strong city did not earn the recommendation.
What does this mean for your business?
It means the field is wide open, because almost no one has built the work AI rewards.
Recommended carries the most weight in the score, so a business AI never recommends cannot leave the Ghosted tier, no matter how clean its site is. And the buyers at stake are valuable: a recommendation inside an AI answer is a pre-sold visit, because the buyer arrives already told your business is the answer.
If 62 of 100 businesses get recommended three times or fewer, the few that get recommended repeatedly take the buyers.
One more detail worth sitting with. Thirty-one of the businesses were never named at all, not even in passing. For those owners, a buyer could ask the exact question their business answers every day and never hear the name.
That is what Ghosted feels like from the outside. Nothing looks broken. The buyer simply goes somewhere else.
What should an owner do with these numbers?
Use them as a map, not a verdict on your business.
The benchmark tells you where businesses like yours tend to fall short. Most are readable and verifiable already, so the highest-return work sits in the middle of the chain. Map the questions your buyers ask. Rewrite your pages to answer them in the first sentence. Add structured data and FAQ content. Then confirm your facts across the web. Those are the Machine-Readable Architecture, Algorithmic Alignment and Targeted Extraction layers of Match Protocol™.
Then measure again. A business that moves from never recommended to sometimes recommended has changed what buyers see first.
How reliable is the data?
The benchmark is directional, not nationally representative.
With n=100, it shows a pattern clearly but cannot speak for every business in America, and the figures describe the small businesses in the benchmark rather than a national rate. Industry and metro groups with fewer than 10 businesses are not reported individually. We published the method so you can see how it was built. The full write-up, with the stage, tier, industry and metro breakdowns, is on the benchmark page.
Frequently asked questions
What does Ghosted mean on the Match Protocol Score™?
Ghosted is the lowest tier, covering scores from 0 to 59, an F on the school grading scale the tiers follow. It means AI rarely or never recommends the business when buyers ask relevant questions. Eighty of the 100 businesses in the 2026 benchmark landed there.
How were the 100 businesses selected?
They are 100 U.S. small businesses with live websites across 10 industries and 11 metro areas in the South, West and Midwest, between five and eleven per group. The benchmark is a fixed panel rather than a random or nationally representative sample, and it will be re-scored with the same method so results can be compared over time.
Can my business beat the average score of 50.9?
Yes. The average is low because few businesses have done the work. Answering buyer questions on the page, adding structured data and FAQ content, and building proof across the web are the moves that lift the Recommended stage, which carries the most weight.
Next step
Find out which side of the 50.9 average you are on. Book Your Match Protocol™ Strategy Session or Get Your Match Protocol Score™.
Clint Rogers is the founder of Modern Asset Lab and the developer of Match Protocol™.
