PELORA Research

Things we measured that nobody else can

PELORA runs live advertising and search accounts in regulated health categories. That produces observations a desk researcher cannot get: what a platform actually refuses, what a category's customers actually type, and where the published rule and the enforced rule come apart. We publish those observations in full, with the method, the raw data, and what we still do not know.

Edited by Preston Durnford, founder, PELORA Marketing Last updated 23 September 2026
Study 01 · Paid search

Google refuses health keywords one at a time, not by page

Two unrelated clinic accounts, same refusal string. Inside one ad group, on one landing page, with one set of copy, Google flagged 2 keywords and passed the rest. That result rules out the explanation almost every agency gives, and it kills a rule we ourselves believed.

23 September 2026 · 2 accounts · 2 flagged, the rest passed Study 02 · Organic search

An industry's own name for itself can be a recruiting query

A home nursing practice held position 6.8 on its category's headline term and converted almost none of it, because the people searching it wanted a job. The phrases buyers use sat at positions 31 to 60, and on the clearest hiring query the practice did not appear at all.

23 September 2026 · 1 account · Search Console plus traced SERPs Tracker · Health ad policy

What Google, Meta, Microsoft, the FTC and the FDA allow in health ads, quoted and dated

A dated, sourced log of health ad rules on Google, Meta and Microsoft, plus FTC and FDA statements, with regenerative medicine first. Last verified 1 October 2026.

2 October 2026 · 24 entries · 5 platforms and regulators Reference · Statistics

Health and wellness marketing statistics, 2026

69 figures a clinic owner or a reporter needs, each read from its primary source on 1 or 2 October 2026, quoted word for word, dated, and followed by what it does not show. Market size, clinic economics, lead costs and ad rules, AI search, brands, home and wealth. No roundup is cited.

2 October 2026 · 69 figures · 66 sources · CSV Review · AI search

How to get your business found on ChatGPT and AI search

We read 20 of the leading studies and platform guides on AI search, from Pew Research Center and the GEO paper to Google, OpenAI and Microsoft. What they agree on, where they disagree, and ten things a business can do.

Sources read 3 October 2026 · 20 sources · 12 figures Evidence review · AI search

What gets a business named by AI: we graded 30 studies

Thirty studies on what gets a business shown in ChatGPT, Google AI Overviews, Perplexity and Gemini, each read at the source and graded on method and commercial interest. One scorecard: what works, what is mixed, what to skip.

4 October 2026 · 30 studies · 7 academic · CSV

Answers

The questions clinics ask us most, answered in full

These are not studies. They are the five questions we are asked most often by clinics whose advertising has been refused, answered from Google's own policy pages with the sub-page number and the date we read it. They are free and they are complete. Nothing is held back for a call.

The standard

How these are written, and what would make us retract one

Marketing content has a credibility problem, and it is self-inflicted. So here is the standard every study on this page is held to, stated plainly enough that you can check whether we kept it.

  1. Every number is transcribed from the account, not remembered. Refusal strings are quoted verbatim from the interface. Position and impression figures come from Search Console, not from a third-party estimate.
  2. The method and the sample size are stated before the conclusion, including when the sample is one account. A finding from one account is labeled as a finding from one account.
  3. What we do not know is published alongside what we do. Every study carries an open questions section. If a pattern fits the data but has no mechanism behind it, we say so rather than dressing it up as a rule.
  4. When the data kills one of our own beliefs, the correction is published. Study 01 does exactly that to a rule we had written down internally and were operating on.
  5. No client is named, and no figure is published that identifies an account rather than carrying the finding.
  6. Contradicting data gets published here with attribution. If you have run the same test and got a different result, send it. That includes, and especially includes, results that make us wrong.

hello@peloramarketing.com

Why this exists

The instrument, not the opinion

Most agency writing is an opinion about marketing. An opinion is not checkable, so it is not citable, so it does not travel. What travels is a measurement with a method and a date on it.

PELORA can produce a narrow class of measurements that most people cannot, because it operates multiple live accounts inside categories that platforms actively restrict. Study 01 exists because two unrelated clinic accounts got refused in the same month and somebody put the two verdicts side by side. That comparison is impossible without both accounts, and it took nothing but writing down what the screen said.

So this page will grow slowly and it will stay narrow. It will not cover topics we have not measured.

Run a regulated account and have a result worth comparing?

We will trade data. If you send a verdict, a refusal string or a query set we can learn from, you get the full working file back and your result published here with your name on it.

Send us a result Book a strategy call