Austin Plumbers AI Visibility Study (October 2026)
Direct answer
On Oct 29, 2026, we asked ChatGPT and Gemini 4 common Austin plumbing questions, 3 times each. The answers were unstable. Only 23% of the business-and-question pairs came up in all 3 runs, and 55% came up just once. The two tools mostly disagreed: of 45 different plumbers named, only 9 were named by both.
What we found
- ChatGPT gave 12 answers with 44 business mentions, about 3.7 businesses per answer, covering 22 different businesses.
- Gemini gave 12 answers with 65 business mentions, about 5.4 per answer, covering 32 different businesses.
- The two tools rarely agreed. Across both, 45 different businesses were named, and only 9 by both tools. On "emergency plumber Austin TX open now," none were named by both.
- Answers changed from run to run. Only 15 of 65 business-and-question pairs (23%) appeared in all 3 runs, while 36 (55%) appeared only once. The broad "best plumber" question was the least stable. The emergency question was the most stable: ChatGPT named the same 4 businesses in all 3 runs, in a different order each time.
- No AI Overview appeared in our single check per question; on 3 of 4, Google showed a loading label but no answer text.
- ChatGPT cited directories. Contractors Ranked (all 3 'best plumber' runs). SiftPros and The Very Best of Texas were each cited in 2 of ChatGPT's 12 answers. Angi appeared in 2 water-heater answers, and Yelp and Best Pick Reports in only 1 answer each. Otherwise ChatGPT mostly cited the businesses' own websites.
- Gemini rarely showed sources. It showed none in 8 of its 12 answers. Where it did show them, they were almost all the businesses' own websites.
- Some local businesses were left out entirely. Of 5 local plumbing businesses we tracked, 3 were never named in any of the 24 answers. The other 2 were named once each, both times by ChatGPT.
Method
- Date: all runs were done on Oct 29, 2026.
- Questions: we asked these 4 exactly as written: "best plumber in Austin TX," "emergency plumber Austin TX open now," "most trusted family-owned plumber in Austin" and "affordable water heater repair Austin TX."
- ChatGPT: we used chatgpt.com logged out, with a fresh chat for each run. Web search ran on its own, and no model name was shown. That made 3 runs per question, 12 answers in total.
- Gemini: we used gemini.google.com logged out. The model shown was "Flash-Lite". We ran each question 3 times in fresh chats, 12 answers in total.
- Google Search: we did a single check per question for an AI Overview (one question was loaded twice).
- Perplexity: it's not included. Logged out, it asked us to sign up instead of answering.
- Counting: when the same business appeared under slightly different names, we counted it as one.
Limits
- Sample size: this is a small snapshot: one city, one trade, 4 questions and one day.
- Versions tested: these are logged-out, free versions only. Signed-in or paid versions may answer differently.
- Location: device location and personalization weren't controlled.
- Source counts: these are minimums, because we didn't open ChatGPT's collapsed source lists.
- Not tested: Google AI Mode.
FAQ
Why run each question 3 times?
AI answers vary. A single run can make a business look visible or invisible by chance.
Why don't you name the businesses?
This study looks at how AI tools behave, not at individual companies.
Will you repeat it?
Yes. We plan to rerun this study quarterly.
Published 2026-10-30 by Answerward.