What Claude finds when someone asks for a plumber in Columbus, Ohio
Claude API with web search, 3 runs, 2026-11-03 21:47–21:52 UTC. 1 of 3 runs searched. A snapshot, not a ranking.
We asked Claude the same question three times on November 3, 2026, through the Claude API with web search turned on. It searched only once, in the second run. Everything on this page about search results comes from that one run. In the other two runs, Claude answered without searching and named no plumber.
- Question, verbatim: "Who is a good plumber in Columbus, Ohio?"
- Model and settings: claude-sonnet-5-5 through the Claude API, effort low, web search tool
web_search_20260209limited to 1 search per answer, dynamic filtering on (the default), max_tokens 4000, no system prompt. - Run times (UTC, 2026-11-03): 21:47:45 (no search), 21:49:36 (searched), 21:51:17 (no search).
- Search Claude ran, run 2 only: "best plumber Columbus Ohio reviews"
Does Claude search the web every time I ask it something?
No. For this question, Claude searched in 1 of 3 runs, with the same question and the same settings each time. Anthropic's web search documentation says "Claude determines when to search based on the prompt." So with search turned on, an answer about local plumbers may still come from the model's general knowledge rather than from the web.
Run 1 said it had not searched. It opened: "I can't recommend a specific plumber in Columbus. I haven't searched for current reviews." Both answers sent the reader to places to look instead: Google Maps, Yelp, Angi, Nextdoor, the Better Business Bureau, and the Ohio Construction Industry Licensing Board to verify a license. Both offered to run a search if the reader described the job.
Across all 10 buyer questions in this study, Claude searched in 10 of 30 runs. The 6 questions it never searched are on the hub page.
Which websites did Claude's search return for plumbers in Columbus?
In the one run that searched, the results held 9 pages on 8 domains. Claude named two of them in its answer: a Yahoo roundup and Angi's plumber list. No plumbing company's own website was in the results.
| Domain | Page title, as the search returned it | In the search results (runs of 3) | Named in the answer (of those runs) |
|---|---|---|---|
| local.yahoo.com | Best plumbers in Columbus, Ohio | 1 | 1 |
| angi.com | 10 Best Plumbers in Columbus, OH - Nov 2026 (Verified Reviews) | 1 | 1 |
| houzz.com | TOP Local Plumbers, Companies & Services in Columbus, OH - Updated August 2025 | 1 | 0 |
| yelp.com | THE BEST 10 Plumbing in COLUMBUS, OH - Updated 2026 - Verified Listings Only - Hours - Yelp | 1 | 0 |
| homeguide.com | The 10 Best Plumbers in Columbus, OH (with Free Quotes) | 1 | 0 |
| yellowpages.com | Best 30 Plumbers in Columbus, OH with Reviews | 1 | 0 |
| entrepreneursofcolumbus.com | Discover the 10 Best Plumbers in Columbus - Local Favorites | 1 | 0 |
| homeadvisor.com | Faucet Install Replace; Repair or Service BBQ Grill (two pages: Columbus, OH and Thornville, OH) | 1 | 0 |
"Named in the answer" means the site's or business's name appears in Claude's answer text. The API returned no citation data in these runs, so neither column measures citations.
What do the listed pages have in common?
Seven of the nine are lists of plumbers. 7 titles say "best" or "top", 7 name Columbus or OH, and 3 carry a year or month, including Houzz's "Updated August 2025". The other two are HomeAdvisor pages for single jobs: faucet installation in Columbus, and BBQ grill repair in Thornville, Ohio.
The same two sites, Yahoo's local roundups and Angi's lists, were the two named in every answer to the roofing version of this question.
What could we measure on those pages?
Only 2 of the 9 URLs served us the page, so there is little to measure here. The Yahoo roundup had JSON-LD (ItemList, NewsArticle, Organization) and 10 headings phrased as questions. The Houzz page had neither. The other 7 answered our server with HTTP 403 or 429 and are not measured.
We fetched each URL from our server on 2026-11-03 between 21:53 and 21:56 UTC, with a Chrome user agent, and read the HTML as served before JavaScript ran.
| Domain | What our server got | JSON-LD types | Question H2/H3 | robots.txt: Claude-SearchBot | robots.txt: ClaudeBot |
|---|---|---|---|---|---|
| local.yahoo.com | page (200) | ItemList, NewsArticle, Organization | 10 | allowed | disallows / |
| angi.com | HTTP 403; not measured | not measured | not measured | allowed | allowed |
| houzz.com | page (200) | none | 0 | allowed | allowed |
| yelp.com | HTTP 403; not measured | not measured | not measured | disallows / | disallows / |
| homeguide.com | HTTP 403; not measured | not measured | not measured | allowed | allowed |
| yellowpages.com | HTTP 403; not measured | not measured | not measured | not measured | not measured |
| entrepreneursofcolumbus.com | HTTP 429; not measured | not measured | not measured | allowed | allowed |
| homeadvisor.com (2 URLs) | HTTP 403; not measured | not measured | not measured | allowed | allowed |
robots.txt allowed Claude-SearchBot on 7 of the 9 URLs. yelp.com's robots.txt disallows / for Claude-SearchBot, with exceptions only for /article pages. Its search page was in the results of run 2; this record does not show how it got there. We could not read yellowpages.com's rules.
Does your homepage title say what you do and where?
Look at your homepage's title, the text in the browser tab or inside <title> in the page source. Check that it says "plumber" or "plumbing" and names your city. 7 of the 9 result titles here named Columbus or OH. The two that did not were the HomeAdvisor job pages.
We did not test whether this changes what Claude finds, and no plumber's own page was in these results to compare against. The free Corpus Plan audit checks page titles and headings as part of five areas: SEO, AI citability, speed, trust and content. It also reads your robots.txt and separates answer crawlers such as Claude-SearchBot from training crawlers such as ClaudeBot. The checks never ask a model.
Free preview with no account · free account for the full report · up to 20 pages per audit.
Run the free audit on your site · How the audit works
What this does and does not show
- One search, not three. Only run 2 of 3 searched, so every count in the tables comes from a single run and can't show repetition.
- One question, one city, one model, one day: three runs between 21:47 and 21:52 UTC on 2026-11-03.
- "In the search results" means the URLs the search returned. Anthropic's documentation says that with dynamic filtering, Claude's code filters the search results before they reach its context window. In run 2, the code tried to print 6 results, hit an error, and then printed the first 3,000 characters of the results. That text held only the first 2 results, the Yahoo roundup and Angi's list, the two sites named in the answer. The record does not show Claude reading the other 7 pages in the table.
- The API returned no citation data, so nothing here measures citations.
- This is the Claude API. We did not test the Claude.ai app.
- We measured a state, not a cause. Nothing here shows that a list page, a title or a robots.txt rule makes a site more likely to be found.
- 7 of the 9 URLs answered our server with HTTP 403 or 429 and are not measured.
- It is not a ranking of plumbers or of websites.
Same study, other trades: web designers · roofers · landscapers · all 10 questions
Related: What ChatGPT and Google cite: three measured studies · Does my website show up in ChatGPT?