On 8 October I published, under Citably, a census of every med spa in the East Cobb zips. Twenty businesses, four local queries, and ten of the twenty came back for none of them. Then I did the thing I should have done first. I ran the same kind of checks on my own site a week later, with no special treatment.
Did the robots.txt setting explain the zeros?
No. All 20 East Cobb med spas measured on 8 October 2026 already allowed the crawlers that fetch pages to cite them in answers, and the ten invisible ones were allowed too. OpenAI’s crawler documentation says OAI-SearchBot “is used to surface websites in search results in ChatGPT’s search features” and that “each setting is independent of the others,” so blocking the training crawler GPTBot does not remove you from ChatGPT search. Google’s documentation says Google-Extended does not impact a site’s inclusion in Google Search. The crawler setting everyone fixes first was not the gap.
My own file passes the same test. The robots.txt on this site allows everything by default and names nine AI user agents with an explicit allow: GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended and CCBot. There is no Disallow anywhere in it. I read the live file on 15 October to confirm. That check is green, and it told me nothing about whether anyone finds me, which is the lesson I keep paying for.
Would my own site have been in the answer?
No. On 15 October 2026 I ran four local queries shaped like the ones in the med spa census, and socialaibusinesssolutions.com came back for none of them. In the census, that is the zero bin, where half the market sat. I also searched my business name with “Marietta GA” in Claude’s web search tool that day. Nine results came back and none was mine. A second search on the bare business name, no city, missed too.
Two honest limits on that table. The first two queries ran through a local search tool that pulls Bing and DuckDuckGo results. Its engine pool then collapsed and returned generic AI-definition pages for the other two, so I reran those with Claude’s web search tool. That is two different surfaces, and the med spa census used a third. A zero here means “not in the results I saw on that day,” the same floor reading I put on the census. I also did not measure whether anyone types “AI consultant 30067” at all. A med spa has local searchers. An AI consultant may not.
What does my site tell an engine about where I am?
socialaibusinesssolutions.com gives engines two locations that disagree. The homepage address block prints a Marietta street address in zip 30067, with an embedded map. The line under it, the hero copy and the footer all say Atlanta. So does the Organization structured data, which carries no street address, and so does the founder’s Person block and the Headquarters line in llms.txt.
The census flagged one med spa whose site never printed a street address on the page I reached, which left engines to infer its location from directories. Mine prints an address and then contradicts it everywhere else. An engine that reads the schema gets Atlanta. A person who reads the address block gets Marietta. As of 15 October 2026 I have not changed any of it. The fix is small: make the PostalAddress, the footer line and the hero copy agree with the address block, Marietta in the address field, with East Cobb kept in the copy and the service area because it is unincorporated Cobb County and has no post office of its own.
What did the census say matters, and where does that leave me?
It found that clean structured data did not get a business picked. OMNI Medspa, about the cleanest markup in the set, came back on none of the four queries. The ones that did come back were present in the directories the answer surface leaned on and matched the query closely. I covered the detail in the full East Cobb med spa index on Citably, with all 20 businesses ranked and linked.
That points at the check I have not done. I have not audited my Google Business Profile. The one directory listing I found in my backlink audit puts the business in Miami, and in that post I called Atlanta the correct city. That was my metro shorthand, and it is the same mistake my markup makes. An engine currently has three cities to choose from.
Here is where it lands. Robots.txt: pass. Local queries: zero of four. Address in markup, footer and hero: a different city from the address block. I had been proud of the first line and had not looked at the other two. It is the trap I wrote up in the four metrics I ignored for my first six months, and the audit that found exactly one inbound link showed the same gap. A green setting is not a result. If you want the same checks run on your site before you trust a green dashboard, that is part of my AI consulting work.