Generative engine optimization · Original data
Directory citations to AI answers held flat for a month. For med spas, they jumped from 37% to 43%.
Since late August we've been running the same GEO probe query set against ChatGPT every few weeks: ten service categories, five cities, questions split across comparison, booking, symptom, and procedure intent. The first time we lined up two months of that data side by side, the headline looked dramatic. Directory citations seemed to be climbing fast. Then we dug into why, and the real cause wasn't the one we expected.
The raw numbers from our August pilot and our September 28 run read like this: 15.8% of citations went to directories in August, 26.6% in September. Nearly a 70% jump in a single month. Before writing anything about it, we checked the cause, and the answer was a little embarrassing. Our directory classifier is a list we maintain by hand, and we'd added dozens of domains to it between those two dates, mostly while fact-checking other posts on this blog. A site cited in both August and September only counted as "known" once we'd actually found and added it. The list grew. What it caught grew along with it. That isn't a shift in AI behavior. That's us catching up on our own homework.
So we reran the comparison properly: the same classifier, the one we use today, applied to both months at once. No domain gets credit for being added later or gets left out for being added earlier. A "known directory" in our data means a site whose homepage, once we actually pulled it up, turned out to be a listing, ranking, or review aggregator rather than a real business. We don't guess from the domain name. Every addition to that list this year came from someone on our side opening the page and checking it by hand.
With that fixed, the real picture is much flatter than the first number suggested. Across all ten categories combined, directory share moved from 32.5% in August down to 30.2% in September. A small decline rather than a climb. Anyone claiming that AI search sends more traffic to directories every month, across the board, isn't describing what we're seeing.
The flat overall number was hiding movement underneath it, though. Broken out by category, most industries barely shifted. Dentist held steady at 31% and 30.7%. Personal injury lawyer stayed close to a quarter, 24.2% to 23.8%. Auto repair, veterinarian, hair salon, optometrist, and gym all dropped, some by a meaningful margin: veterinarian from 27.1% to 22.4%, hair salon from 34.9% to 29.1%, optometrist from 36.6% to 30.7%.
Two categories went the other way. Plumber ticked up slightly, 27.5% to 29.1%. Med spa jumped from 37.2% to 43.0%, the single biggest move of any category in either direction, and the only one that climbed by more than two points.
Med spa is also the category we've covered the most on this blog, since it's the one we built askChat around first. We've already documented a cluster of aesthetics-specific directories that surfaced in earlier citation audits and weren't on anyone's radar before: InjectorIndex, ToxAura, MedSpaScout, USMedSpaDirectory, TheClinicCompass. These sites exist because medical aesthetics is a high-value, high-search-volume field, the kind that draws SEO-built listing pages in a way plumbing or pet care simply doesn't. If that cluster of aggregators is still actively publishing and getting crawled, a one-month jump in how often they get cited makes sense. It's the category where this kind of competition pays off the most.
It also lines up with something we found in a different cut of this same data a few weeks ago. For med spa, directory dominance hinges heavily on what someone actually asks. A rating question like "best med spa near me" pulled a directory most of the time, while a symptom question like "something for acne scars" almost always named the clinic directly. Our twenty questions per category stayed identical between August and September, so a shift toward more comparison-style phrasing can't explain this. The same directories are simply winning more of the same questions they were already competing for.
We want to be upfront about what a finding like this can and can't support. These are two single-day snapshots, three and a half weeks apart, through one query set and one AI engine. A different week could move any of these numbers by several points in either direction, and we have no way yet to know whether med spa's jump marks the start of a pattern or reverses on the next run. What we can say with more confidence is the overall-flat result, since it held across nine of ten categories and ran opposite to the number we nearly published by mistake.
For a med spa owner, the practical takeaway isn't "panic, directories are winning." It's narrower: this is currently the one category in our dataset where directory citations are moving fastest, and the businesses that keep showing up in the 57% that still goes to an actual clinic page tend to have treatment-specific pages instead of a homepage that lists services in general terms. If a directory with your treatment and your city in its title is outranking you in ChatGPT's citations today, that's worth checking directly rather than assuming the same pattern applies to every type of local business.
We'll rerun this same comparison next month with the same classifier, and we'll publish the next data point whether it confirms this one or not.
Source: askchat.studio's GEO probe database, directory-citation share across ten service categories (med spa, veterinarian, personal injury lawyer, chiropractor, auto repair, optometrist, plumber, gym, hair salon, dentist), five U.S. cities, twenty queries per category spanning comparison, booking, symptom, and procedure intent. August pilot collected August 26-27, 2026; follow-up run collected September 28, 2026. Both periods reclassified with the same directory list in use as of this post, through the OpenAI Responses API with web search grounding.