Field research · Auckland · June–July 2026

The agents aren't
being blocked.
They're being ignored.

A popular story says the web is walling itself off from AI crawlers. Among Auckland's small businesses, we found the opposite: the doors are wide open. Openness just turns out not to be the thing that gets you recommended.

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.

You've read the headlines: publishers blocking GPTBot, robots.txt as the new battleground, the web going dark to AI. That may be true of large publishers. It is emphatically not what small-business websites look like from an agent's point of view.

  • Of 679 probed Auckland businesses, just 28 deny any AI crawler in robots.txt.
  • When we actually fetched each site the way an AI crawler would, only 19 blocked the request.
  • That's roughly 96% of small-business sites fully open to AI systems — across restaurants, barbers, plumbers, and retail alike.

The usual caveats apply: this is a June 2026 snapshot with one crawler user-agent tested per site — policies change, and a different bot might be treated differently. And the sample is small-business websites; large publishers blocking AI crawlers is real and well documented. This finding doesn't contradict that story — it just shows the story doesn't describe main street.

So when 60% of these same businesses never surface in any AI engine's answers (the invisibility wall), access isn't the explanation. The engines could read almost everyone. They simply have no reason to: nothing they trust has written those businesses down.

Why this matters for the advice you're given

A lot of "AI readiness" consulting is access-flavoured — crawler permissions, technical SEO for bots, feeds and files. Necessary hygiene, occasionally decisive, but in our data the blocked-door case is rare (and sometimes accidental — a security service swatting bots the owner never configured). The binding constraint is upstream: authority, not access. An open door on a street nobody walks down.

One nuance we did find worth checking on your own site: readability failures that look like content but read as nothing. About one site in eight across our restaurant runs was a JavaScript shell serving a nearly-empty page to a plain fetch. Those aren't blocked — they're mute. Different problem, same result.

What to do about it

Spend five minutes on hygiene: confirm your robots.txt isn't accidentally denying AI crawlers, and that your key pages render actual text without JavaScript. For roughly 96% of you, this is already fine.

Then stop optimising the door and work on the street. Citations in sources engines trust are what generate visits from agents — that's where the invisible 60% and the visible 40% actually differ in our data.

Open door, empty street?

A shelf check tells you which problem you actually have — access, readability, or authority — instead of guessing.

Request a shelf check →