ARA Oral Care Study — Q3 2026. Brand Intelligence for the oral care category, using the four frontier AI models as the shopper.

462 choices / 4 models x 15 buyer questions / audited brands in colour, every other brand in a stepped grey / naming variants merged before counting.

Ask a machine which toothpaste to buy and it will name one hundred and thirty-five different brands.

In the third quarter of 2026 we put fifteen shopper questions to ChatGPT, Claude, Gemini and Perplexity — the best everyday toothpaste, what to use for sensitive teeth, what dentists actually recommend, what to do about chronic bad breath, what is worth paying more for — and recorded every brand named in every answer, and where it ranked. Four hundred and sixty-two choices in all.

The eleven brands we audit, which between them hold most of the physical shelf in most of the world, hold 41.6% of this one. Colgate takes 7.8% of it, Crest 7.4%, Sensodyne 5.2%. For scale: in women's deodorant, one of the tightest shelves we have charted, three brands hold 39% between them and the leader alone takes 13.8%. Oral care is not that. Nor is it travel, where 187 brands appear and none reaches 4%. It sits between the two, and it is moving in one direction.

The interesting ground is not the tail — it is the middle. Boka, TheraBreath, Davids, Philips Sonicare and RiseWell were each chosen by all four models, and each of them can take first place on a question. Every one of them outranks Arm & Hammer. Every one outranks Parodontax. All five outrank Marvis and Quip together. These are not fringe names on this shelf; they are mid-shelf brands sitting directly among the majors, and most executives in the category could not name three of them.

Then the tail, and it is long. Eighty-six brands were named exactly once, and between them they account for 19% of the whole shelf. Another seventeen were named twice. That is not noise. It is what a shelf looks like before it settles.

Split the shelf by question and the shape becomes plain. On the eight questions carried word for word from April, the eleven audited brands hold 48.2%. On the seven newer ones — kids, bad breath, what is worth paying more for, what belongs in your bathroom — they hold 33.6%. The older the question, the better the old names do.

Measured on those eight identical questions, every one of the eleven lost share this quarter. Not one gained. Between them they went from 63.1% of those answers in April to 48.2% now.

One caveat, because it cuts against our own headline: this is the first oral care reading in which all four models searched the live web, and answers ran deeper — 7.8 brands named per answer against 6.4 in April. Better-informed models find more real brands, so some of that challenger gain is the instrument improving rather than the incumbents slipping. Read the direction, not the decimal. The direction is not in doubt.

Last quarter this shelf looked settled, and we said so: a top three that had barely moved in thirty years, and challengers who had built beautiful products a machine could not verify. One quarter later the same eleven brands hold two thirds of what they held, and the ground they gave up went to brands that can answer a specific question. The shelf did not shrink around them. It got longer, and they did not grow with it.