Field research · Auckland · June–July 2026

60% of real businesses
don't exist to AI.

We built exhaustive lists of every Italian, Japanese, sushi and Indian restaurant, every barber, and every board-game shop in Auckland — then asked the AI engines about each vertical, every way a customer would. 321 of 537 businesses never came up. Not once, on any engine.

SCOPE — Auckland, New Zealand · June–July 2026 snapshot. 7 verticals, 679 probed businesses: Italian, Japanese, sushi and Indian restaurants, barbers, plumbers, and board-game shops. Engines queried via their APIs — Gemini with Google Search grounding, OpenAI web_search, Claude with live search, and Apple Maps (MKLocalSearch, the layer Siri uses for local queries). APIs are clean, reproducible instruments and proxies for the consumer products, not the apps themselves. One city, one snapshot — treat this as field data, not a global law.

The premise of this research was simple: if AI agents are becoming a way people discover local businesses, someone should measure who the agents can actually see. So we did it properly — not "ask ChatGPT and screenshot the answer," but building a verified master list of every business in a vertical (cross-referenced against the Apple Maps POI layer), then sweeping four engine families with the queries a real customer would use, and checking which of those real businesses ever appeared.

The headline: pooled across our six exhaustive verticals, 321 of 537 businesses (60%) were invisible to every AI engine we tested — no mention, no citation, nothing, across every query and every model tier. If anything that's an undercount: our name-matching is deliberately generous, so a business named after its category ("Sushi Takeaway") can be credited as seen when the answer merely used those words. Each query ran once per tier — a zero across four or five queries and every model is meaningful, though it isn't infinite sampling. And a business missing from the Maps layer entirely never made our master list in the first place.

VerticalBusinessesInvisible to every engine
Japanese restaurants12279
Sushi11166
Indian restaurants10874
Barbers10461
Italian restaurants7339
Board-game shops192

Two readings of that table matter. First, the wall is consistent: every food and service vertical loses half or more of its real businesses. Second, the one exception is instructive — board-game shops are a tiny, well-documented market (19 shops, a passionate community writing about them), and almost all of them surface. The wall isn't about AI disliking small business. It's about how much of the web talks about you.

The long tail exists only on the map

Most of the invisible businesses do exist somewhere an agent could theoretically reach: they're in the Apple Maps layer, which is why our master lists could include them. In the restaurant verticals, Apple Maps (the layer Siri draws on) actually surfaced more individual venues than any language-model engine — 25 to 30 per vertical in our runs. But being a pin on a map is not the same as being recommendable: a pin gets you found when someone searches your name, not chosen when someone asks "where should we eat Friday?"

And before you assume the invisible 60% are blocking AI crawlers or have broken websites: they aren't, and mostly they don't. Only 28 of all 679 probed businesses deny AI crawlers in robots.txt — we wrote that up separately in Nobody is blocking the agents. The wall is an authority problem, not an access problem: nobody independent writes about these businesses, so engines that assemble answers from written sources have nothing to assemble from.

What to do about it

Check which side of the wall you're on. Ask an AI assistant the question your customers would ask — not your name, the need ("best Japanese in Auckland for a date"). If you never come up, you're in the 60%.

Then fix authority before aesthetics. The invisible businesses aren't invisible because their sites are ugly — they're invisible because no list, review platform, or community thread mentions them. Getting written about is the way through the wall. Where those citations live per vertical is the subject of Every vertical has a kingmaker.

Which side of the wall are you on?

We run this exact measurement for individual businesses — every engine, every query your customers actually ask, verbatim results.

Request a shelf check →