How often does an AI name a local business?

Across 25 trades and 66 US metros, 79% of the 29,486 local businesses we checked were never named when we asked leading AI assistants who to call, in checks run between July 4, 2026 and October 3, 2026 (newest run per trade and metro; see “How we counted this” below).

Some checks named nobody at all

Out of 536 trade-and-metro audits in this edition, 29 came back with zero businesses named. In each of those, every question we asked in that trade and metro got an answer from AI that didn't name a single business on our list.

Chimney Sweeps: this trade

For chimney sweeps, 48% of the 435 businesses we checked across 12 metros were never named by ChatGPT in this edition.

All trades

Every trade we benchmark, sorted by never-named rate, highest first.

How we counted this

Every trade-and-metro board in our network is measured the same way: we ask AI who to call, the way a customer would, several times with different wording, and record every business it names. Full details, including how the question pool and the businesses checked are built, are on our methodology page. A few of our trades have run their benchmark more than once. To avoid counting the same business twice, this page applies one rule everywhere: We keep the newest “updated_at” row for each trade in each metro (matched on trade, city and state), and drop any older repeat.

Limits

• Which AI we measured. Most of these checks were run on ChatGPT, using OpenAI's gpt-4o model with web search turned on. A handful of our earliest editions didn't record which AI system answered, which is why this page describes the measurement in general terms as “AI” rather than naming one product for every trade. A different AI tool, like Google's AI answers, may name different businesses than we found.
• Each check asked several questions, not one. A trade-and-metro audit is 4 differently worded questions in most editions (8 in a few, where every question was asked twice). An AI's answer can change from one run to the next, so “never named” means never named across the questions we asked in this edition. It does not mean the business could never be named.
• The pool comes from Google Maps. For each metro, we built the list of businesses to check from a Google Maps search for that trade. A real business that doesn't show up in that search was never part of the pool and was never checked, named or not.
• How duplicates were removed. We keep the newest “updated_at” row for each trade in each metro (matched on trade, city and state), and drop any older repeat. Some of our snapshot files also store a pre-added-up total for a whole trade so we don't count those totals a second time on top of the metro rows (see “How we counted this” above). Skipping the de-duplication step entirely and counting every row as it sits in our files puts the never-named rate at 79.2% instead of the published 79.0%, close to the published number and not a different story.
• Edition date. This page reflects benchmark runs between July 4, 2026 and October 3, 2026. Each trade shows its own newest run; the network number is a mix of editions run on different dates.

Cite this

"79% of 29,486 local businesses checked across 25 trades and 66 US metros were never named by AI." Source: https://mostrecommendedchimneysweeps.com/ai-named-rate (edition dated October 3, 2026).