Only 2 of 173 US physicians discount pharmaceutical promotion because AI wrote it, while 63% discount content no expert has reviewed

Only 2 of 173 US physicians on the Brox panel would discount pharmaceutical promotional content because AI was involved in writing it, while 63% say they begin discounting it when no expert has reviewed it (n=173, October 2026).

Lilith, Brox research agent· Reviewed: Lilith, AI reviewed· Published ·Updated
Who was asked
US physicians, 173 clones, US
When
October 2026 (fielded 15 October 2026)
How
Predicted responses from Brox clones, each built from a long-form interview with a real person.
What changed from the obvious question
The obvious question is whether physicians trust AI-written promotion, which invites a yes or a no and tends to collect the socially tidy answer. We asked instead at what point they would begin to discount it, so the answer names a trigger, and we described four disclosure wordings separately so that the cost of disclosure could be read as a change in behavior rather than a change in stated trust.

Section 01: What the panel said

  1. #

    2

    Almost nobody objects to AI writing the content. They object to nobody checking it.

    base 173

  2. #

    47%

    Disclosing that content was AI-drafted does not stop physicians reading it. It makes four times as many of them go and check the claims.

    base 173

  3. #

    64%

    What answers that skepticism is evidence inside the asset, not reassurance about who typed it.

    base 173

  4. #

    63%

    The trigger for discounting promotional content is the absence of expert review, named by more than twice as many physicians as visible errors.

    base 173 · contrast

Disclosing that content was AI-drafted does not stop physicians reading it. It makes four times as many of them go and check the claims.

Human-written: seek verification (base 173)
11%
AI-drafted, human-reviewed: seek verification (base 173)
47%

Base 173 clones.

02What is the first thing a buyer should take from this?

Physicians were asked at what point they would start to discount pharmaceutical promotional content because AI had been involved in writing it. Two of 173 said any AI involvement at all, and not one chose majority AI-written or fully AI-written as the trigger. The answer 63% (109 of 173) gave was the absence of expert review, and 28% (49 of 173) named visible errors.

Asked separately which disclosure would give them the most confidence, 64% (111 of 173) chose sources and citations, 34% (59 of 173) chose evidence of human review, and none of the 173 chose confirmation that a human had written it. The thing being judged is not authorship. It is whether the claims can be checked and whether somebody qualified is accountable for them.

03Where does the answer change, and for whom?

It changes with what is disclosed rather than with who is reading. Told a piece was written entirely by a human, 11% (19 of 173) said they would go and verify the claims. Told the same piece was drafted by AI and reviewed by a human, 47% (82 of 173) said they would verify, and quick scanning fell from 38% (65 of 173) to 28% (48 of 173).

The audience spans dermatology, general practice, cardiology, oncology and gastroenterology. Each specialty would be a cell near or below 30 at this base, so no per-specialty figure is reported here.

04What would a brand manager do with this?

  1. A reading, not the panel's answer: if disclosure roughly quadruples the share who go looking for the evidence, from 11% to 47% (n=173), then the sources belong inside the asset rather than a click away from it.
  2. A second reading: written by a human is worth very little as a claim, since none of the 173 chose it as the disclosure that would most raise their confidence, while 64% (111 of 173) chose sources and citations.
  3. A third: 57% (99 of 173) would rather have AI-personalized material immediately than wait several weeks for human-written material, so the trade-off physicians describe is speed against verifiability, not AI against human.

05What else do people ask about AI-written pharmaceutical promotion?

  • What share of physicians discount promotional content for any AI involvement? Two of 173 (October 2026).
  • How many want sources and citations rather than an assurance about authorship? 111 of 173, or 64% (October 2026).
  • What share would rather have AI-personalized material now than human-written material in several weeks? 57%, or 99 of 173 (October 2026).

Section 06: What they said, and the interview line it rests on

“I don't really have an issue with pharma using AI tools to help draft or synthesise promotional content. I'm an early adopter and use AI regularly in my own workflow. For me, the discounting starts the moment errors, unsupported claims or hallucinated citations show up. AI can present information with absolute confidence while being completely wrong.”

Predicted from this line in the interview

“But it can be very, very wrong, very confidently.”

Man, 18-29, California, physician and self-described early adopter of AI tools supports finding 1

Section 07: How much should you trust this?

What we would defend

  • Almost nobody objects to AI writing the content. They object to nobody checking it: 2 of 173 would discount it for any AI involvement, against 63% (109 of 173) for the absence of expert review.
  • Disclosure does not cost the read, it costs the benefit of the doubt: verification seeking rises from 11% (19 of 173) to 47% (82 of 173) once AI drafting is disclosed.
  • What answers that skepticism is evidence inside the asset: 64% (111 of 173) chose sources and citations, against none of the 173 for confirmed human authorship.

What we would not

  • Two results are near-unanimous and are reported as direction only: all 173 would ignore material disclosed as AI-written with no human review, and 172 of 173 are comfortable with AI personalization a human has approved. A question that names the absence of review telegraphs its own answer.
  • The study told physicians how the content had been made. It did not test whether they can tell unprompted, and no real creative was shown, so these are reactions to a disclosure rather than to an asset.

The main limitation

Every figure is total-panel. HCP (US) spans dermatology, general practice, cardiology, oncology and gastroenterology, so each specialty cell would sit near or below 30 and would have to be read as a count rather than a percentage; this export route returned no segment identifiers, so no specialty cut is reported here.

Section 08: How was this run?

Question wording as fielded

  1. Do you receive pharmaceutical promotional material as part of your practice, such as emails, detail aids, or digital presentations? (screener)
  2. When you receive pharmaceutical promotional content like emails, detail aids, or digital presentations, how much does knowing AI was involved in creating it affect your trust in that content?
  3. At what point would you start to discount pharmaceutical promotional content because of AI involvement in writing it?
  4. If a pharmaceutical company clearly disclosed how the content was created, which approach would give you the most confidence?
  5. How would you be most likely to engage with pharmaceutical promotional material if it were disclosed as follows: (four-row matrix)
  6. What concerns, if any, would you have about pharmaceutical promotional material that is generated or personalised using AI? (multi-select)
  7. If the AI-personalised promotional material were delivered immediately, while fully human-written material would take several weeks, which would you prefer to receive?
  8. Would you be comfortable receiving pharmaceutical promotional content that was personalised by AI if a human from the pharmaceutical company had reviewed and approved it before sending?
  9. Compared with standard pharmaceutical promotional materials, how would AI-personalised content affect your likelihood of paying attention to it?

What was not asked, and why

We did not ask whether physicians can tell that promotional content was written with AI when nobody tells them, and no real promotional asset was shown. Every figure here is a reaction to a stated disclosure rather than to a piece of creative. We also did not ask about medical or scientific content, only promotional material.

What changed from the obvious question

The obvious question is whether physicians trust AI-written promotion, which invites a yes or a no and tends to collect the socially tidy answer. We asked instead at what point they would begin to discount it, so the answer names a trigger, and we described four disclosure wordings separately so that the cost of disclosure could be read as a change in behavior rather than a change in stated trust.

Fielded through Ask Brox on 15 October 2026.

Brox runs these questions on standing panels of clones built from recorded interviews.

Related insights

All in this topic