A parent asks ChatGPT which treatment center to call for their son’s alcohol rehab, mentions their Blue Cross Blue Shield plan, and gets back two or three names, not five. In a 12-metro test run on October 1, 2026, that insurance-specific question made ChatGPT cite an insurance company’s own provider page in 10 of 12 answers. A treatment center’s own website was the cited source in only 3 of those 12 answers. If your insurance page cannot answer “do they take my plan” in one sentence, the insurer’s directory is answering it for you, and your center is not in the room.
Run your own center through a free AI visibility scan and see which sources are answering for you right now, before you read the rest of this.
What the test actually asked, and what came back
The study put two questions to ChatGPT, logged out with web search on, across 12 US metros (Atlanta, Boston, Chicago, Dallas, Denver, Houston, Los Angeles, Miami, Nashville, Philadelphia, Phoenix and San Diego): a general “best drug rehab in [city]” question, and a parent’s question naming Blue Cross Blue Shield and alcohol rehab for a son. The general question returned five names on average, each with a disclaimer that the right fit depends on level of care. The insurance question changed the shape of the answer entirely: shorter lists, a median of two to three centers, and in two cities a single center named to call first.
That shift matters because of where the citations came from. Across the 12 insurance answers, an insurance company’s own page (most often the Blue Cross Blue Shield Association’s Blue Distinction Centers list for substance use treatment) was cited 10 times. A treatment center’s own website was cited in just 3 answers, and in those 3 cases it was because the site stated plainly, in a sentence, that it accepted the plan.
| What the parent’s ChatGPT answer contained | How often (of 12 answers) |
|---|---|
| Told the family to verify their exact plan | 12 of 12 |
| Cited an insurance company’s own page | 10 of 12 |
| Cited a treatment center’s own website | 3 of 12 |
| Warned about dangerous alcohol withdrawal | 8 of 12 |
| Gave a phone number to call | 6 of 12 |
Why a logo is not a citation
The same researchers read the homepages of 65 treatment providers who ranked for “drug rehab [city]” in those 12 metros, checking how each one presented a Blue Cross Blue Shield plan. The split was stark: 29 of 65 homepages did not mention the plan at all, 18 mentioned it only in a logo image’s alt text, and 13 only in a menu or list of insurer names. Just 5 of 65 stated acceptance in an actual sentence, something like “we are in network with Anthem Blue Cross for detox and residential care.”
That is the difference ChatGPT is reading. A logo strip is built for a human scanning a page with their eyes. A plain sentence naming the plan, the levels of care it covers, and how to verify benefits is a claim a language model can lift and quote. Of the two treatment center citations that made it into a Boston answer, both were the center’s own site, because both said outright, in words, which plan they accepted and for what.
The levels of care gap is where most centers lose the citation
Saying “we accept Blue Cross” is not enough on its own. The study found the centers that got cited specified which levels of care the plan covered: detox, residential, PHP, IOP. A plan that covers residential treatment may not cover sober living, and an insurer directory listing will not make that distinction for you. A center’s own insurance page is the only place that can say, in one place, “in network with [plan] for detox and residential; PHP billed separately; call admissions to verify outpatient coverage,” and that specificity is exactly what gets paraphrased into an AI answer instead of a generic “accepts most major insurance plans” line that a model has nothing to quote.
What to put on the page this month
| Current state | What to change it to |
|---|---|
| Insurer logo strip, no text | One page per major plan, with a sentence stating in-network status and which levels of care it covers |
| “We accept most insurance” | Name the specific plans and the specific levels of care for each |
| No link from the homepage to the insurance page | A visible homepage link, since most centers already have the page, just not surfaced |
| Insurer directory listing out of date | Confirm your center’s entry in the insurer’s own provider directory, since that page gets cited directly |
“The directories and insurers are already answering for you, whether the information is current or not,” said Ryan Rivard, founder of BSPKN. “The fastest fix is rarely new content. It is rewriting the page you already have into a sentence a model can quote.”
None of this requires new content volume. It requires one afternoon rewriting the insurance pages you already have into sentences instead of logos, and a phone call to confirm your insurer directory listing is current. That is also the fastest way to improve how you show up on a free AI visibility scan, which checks these exact citation patterns against your own site and your competitors in your metro.
FAQ
Does ChatGPT actually read my insurance page?
When web search is on, ChatGPT rewrites a question into search queries and builds its answer from the pages those queries return, your site included if it is indexed and your insurance page answers the question directly. The 2026 test found a center’s own site was cited in 3 of 12 insurance-specific answers, each time because the page stated plan acceptance in a sentence rather than a logo.
Why does ChatGPT cite the insurance company instead of my center?
Insurer provider directories, like the Blue Cross Blue Shield Association’s Blue Distinction Centers list, are structured, current and written in plain text that states exactly which centers are in network for which service. That is easier for a language model to quote than a homepage logo strip, so the insurer’s own page gets cited instead of yours.
What levels of care should an insurance page specify?
Detox, residential, partial hospitalization and intensive outpatient are frequently billed and covered differently by the same plan. Stating which of these your in-network status actually covers, rather than a blanket “we accept this plan,” is what let the cited centers’ pages get quoted directly in the 2026 test.
Does this replace needing to be in the insurer’s own directory?
No. Insurer directories were cited in 10 of 12 insurance-specific answers, more than any other source, so keeping that listing current matters as much as your own site. Both need to say the same thing about the same plan.
How do I see how my center shows up right now?
Run the free Peek AI visibility scan. It checks what ChatGPT, Perplexity and Google’s AI features currently surface for your center and your competitors in your market, insurance pages included.
What this means for your program
This is one data point inside a larger pattern BSPKN tracks across healthcare, construction and financial services marketing: the pages that get a brand cited by AI search tools are rarely the ones with the most writing, they are the ones that state a plain fact in a sentence a model can lift whole. If you want a second set of eyes on where your center’s citations are actually coming from, and where the gaps sit against competitors in your metro, book a 15 minute strategy call. See the data behind our Propel OS approach to AI visibility, or read how it has shown up in client results. See also our broader GEO marketing approach and how it applies to recovery and behavioral health marketing.
Source: KhanWork test of ChatGPT (logged out, web search on), 12 US metros, two questions each, insurance-specific question tested against Blue Cross Blue Shield, and a census of 65 treatment provider homepages, October 1, 2026.
