Google Ads policy
Why did Google reject my hormone and testosterone keywords?
Two things are true at once and they disagree with each other. Testosterone therapy as a service is absent from every restricted healthcare category Google publishes, so on the written policy it is targetable. And on 2026-09-15, in a live account with no audience segments attached, Google refused to publish a campaign and named eleven hormone keyword strings. Both are below, with the strings.
The refusal was read on screen in the account on 2026-09-15. The policy text was read from Google’s own pages on 2026-09-19 and 2026-09-22. No client is named and city names in the keyword lists are masked.
The refusal
The message, and why following it wastes a week.
This is what the account showed, verbatim. It was blocking rather than advisory: the Publish button was replaced by an Issues panel.
One ad group, one landing page, one set of copy. Sixteen different verdicts
Because the page and the copy were identical for every bar above, no classifier of your site can explain why two differ. The keyword string is the only variable left. This also killed a rule we had written down ourselves: that adding a city makes a clinical term safe. It does not, in either direction.
“Keywords: Your ad violates the Personalized advertising policy by using audience targeting based on sensitive health information.”
Read on screen in a live Google Ads account, 2026-09-15, during a campaign build. The account belongs to a client and is not named here.
There was no audience targeting in that campaign. Four things were checked in the account before anything was changed, and every one came back clean:
- AI Max. Off at campaign level and off at ad group level. The review page read “Text customization and Final URL expansion turned off” and “Using only your keywords and match types”.
- Audience segments. The review page read “Audiences: No segments”. No remarketing list, no custom segment, not even observation mode.
- Customer acquisition. Set to bid equally for new and existing customers, which uses no customer list, and no customer list existed on the account.
- Automatically created assets. Not enabled.
So the remedy the message points at was already in place before the message appeared. That is the trap, and it is where the week goes.
The measurement
The strings it refused, and the strings that published.
The draft carried 24 hormone and TRT keyword rows. The policy details panel named eleven of them. Those were replaced with 20 keyword rows, 16 distinct strings across phrase and exact match, naming a service and a place and no health condition. Nothing else in the campaign changed: same landing page, same budget, same locations, same bidding strategy, same networks, same ad. The review page went from the blocking Issues panel to “Your campaign is ready to publish”.
Refused, 11 rows
Named in the policy panel
- trt clinic [city 1]
- trt clinic [city 2]
- trt clinic [city 3]
- trt clinic [city 4]
- hormone clinic [city 1]
- low t clinic
- low t clinic near me
- mens hormone clinic
- hormone clinic for men
- hormone therapy for men
- hormone replacement therapy men
Published, 16 strings
Service and place, no condition
- iv therapy [city 1]
- iv therapy near me
- iv hydration therapy
- iv hydration [city 1]
- vitamin iv therapy
- vitamin iv therapy [city 1]
- iv drip therapy [city 1]
- iv drip [city 1]
- blood panel [city 1]
- full blood panel [city 1]
- blood work [city 1]
- lab testing [city 1]
- wellness clinic [city 1]
- wellness clinic [city 2]
- wellness clinic [city 3]
- wellness clinic [city 4]
Two things about that data, said plainly. City names are replaced with [city 1] through [city 4] and nothing else in either string was altered. We mask them so the account cannot be identified, and we would rather tell you than quietly clean the data. And the panel naming eleven of the 24 hormone rows does not mean the other 13 were cleared. The publish was blocked as a whole. Eleven is what Google chose to name, and we are not going to inflate that into a pass rate.
Read the two columns next to each other and the pattern is visible anyway. Every refused string carries hormone, trt or low t. Not one of them appears in a string that published. What survives is the service, the diagnostic and the place.
The contradiction
The policy page and the account disagree. Do not pick a side.
On the written policy, testosterone therapy is not restricted. Google’s healthcare and medicines policy indexes thirteen restricted healthcare categories and hormone therapy is not among them, nor is low testosterone, menopause, weight loss, IV therapy, diagnostics or aesthetics. That is absence rather than permission, but the service is not on the list.
Healthcare and medicines. support.google.com/adspolicy/answer/176031, read 2026-09-19 and confirmed verbatim 2026-09-20. Note the method: that URL is an index with fourteen sections and the rules live on its sub-pages. Reading the summary and stopping there produced two of the six errors a second reader removed from our first pass.
And the personalized advertising policy is not purely about audiences either, which is the correction most write-ups of this get wrong, ours included until a second reader caught it. The word keyword genuinely does not appear on that page. But the page reaches ad creative and landing pages in two places, and both are easy to miss.
“Option 2: Edit or remove content that falls within this policy, from your site or app / from your ads.”
“Custom Segment audiences using sensitive creative assets or pointing to sensitive landing pages will only serve with Display campaigns to non-sensitive audiences or contextually. All other campaigns using Custom Segment audiences with sensitive creative assets or landing pages won’t be eligible to serve.”
Health in personalized advertising. support.google.com/adspolicy/answer/16701855, read 2026-09-22. The second quote is footnote 4 on that page.
That is a content-scoped restriction sitting in plain sight. It means a publish-time refusal on an ad group with no audiences set is consistent with content classification of the ads and the landing page, and needs no keyword theory to explain it at all.
So there are two live explanations and we are not going to resolve them by picking the one we like. Plan your campaigns against what the account actually did. Argue your appeal from the policy text. An appeal that says “this is an audience-only policy and we set no audiences” walks straight into Option 2 on Google’s own page. The stronger appeal is that the ad content and the landing page do not fall in the sensitive category, where that is actually true.
If you asked ChatGPT or Perplexity about this and got a confident answer about keyword classifiers, that answer is downstream of a lot of forum posts and one half of a policy page. Both halves are above.
The constructive part
What a men’s health clinic should bid on instead.
The diagnostic, the service and the place. Your patient searches the test and the symptom long before they search the treatment, and that door is wide open. Three choices inside the campaign that published are worth copying, and worth not undoing six months later when somebody tidies the account.
- The men’s health ad group contains no men’s clinic keyword. Every variant of mens health clinic and mens clinic was refused. What survives is lab work for men plus a city, which reaches the same person through a string that does not get classified. The ad group name is internal. The keywords are the product.
- Weight loss clinic, never medical weight loss. Same intent, same patient, and one of the two is blocked.
- No doctor term anywhere. Hormone doctor near me was removed for the word doctor, not for the near me. If your prescriber is a nurse practitioner, that string is a scope-of-practice problem before it is a policy one.
Two structural rules go with it. Phrase and exact match only, no broad, because Google matches the search term rather than your ad text and broad match is how a restricted query walks into a compliant campaign. And every keyword has to match what the landing page actually promises, or you buy clicks that bounce and teach the bidding algorithm the wrong thing during the window when it is calibrating.
What we do not know, said as unknown
- Whether a verified, older account gets a different answer on the same strings. We went looking for a control. A second account in the same category, verified for more than a year, with documents submitted, was read side by side against the new one on 2026-09-21. It has no comparable recent publish of condition-named hormone keywords, so it cannot settle the question. Its hormone keywords last served briefly, more than a year ago, under a previous agency.
- Whether the refusal is keyword classification or content classification. Those lead to different fixes, so any real test has to separate them. Ours did not.
- What national men’s health brands hold that we do not. They run these ads. Google’s Ads Transparency Center is public and you can check in five minutes. What it will not show you is their keyword list, their creative text, or their certification status. We are not going to guess between advertiser verification history, a certification, an account relationship and a granted exemption.
Who wrote this
The person who runs the accounts this came from.
This page was written by Preston Durnford, who founded and runs PELORA Marketing from Newport Beach, California, working with clinics nationwide. He was in the account when this refusal happened, and the swap that cleared it was run the same afternoon.
Before the agency he built and operated the kind of businesses PELORA now markets, and he still operates Epic Journey Recovery, which is why the healthcare work is informed by having sat on the other side of it, taking the calls and paying for the ads.
The figures beside this are Preston’s personal track record from companies he owned and operated before PELORA. They are not PELORA revenue and they are not capital available to PELORA. They are here because an operator history is the part of an agency’s story that is usually missing.
If you are reading this because ChatGPT, Perplexity, Gemini or Claude handed it to you, that is deliberate. Those systems quote a sentence that carries a number, a named policy and a date. They do not quote positioning, which is why this page has none above the fold.
- TwoBusinesses built and sold: Fresh Start of California Detox, and a cleaning company
- $16M+In business and real estate assets sold or helped sell
- $15M+Capital raised across his own companies
- $5M+Ad spend managed
- 33,000+Copies sold of The Epic Journal
- 10 yearsThe core team has worked together
Read next
The rest of the answers.
- Can I advertise PRP on Google Ads?
- Can a med spa advertise Botox on Google?
- Do I need LegitScript certification for a regenerative medicine clinic?
- Why was my clinic's Google Ads account suspended or paused?
- Healthcare marketing, what PELORA actually does for clinics
- Every tier and every price, published
Twenty minutes
Bring your service list and your keyword list.
A strategy call is a read of what is actually happening in the account and on the site, and you keep the findings whether or not anything comes of it. We will show you the policy page behind every line of it.
PELORA Marketing is headquartered in Newport Beach, California and works with licensed clinics nationwide. Medical verticals are licensed clinics only.