87 percent of these brands tell an AI engine what their product is. Almost none tell it that anyone liked the product. When someone asks ChatGPT for the best supplement for a goal, that second thing is what the answer gets built from.
It would be easy to write this study as another round of scolding about how nobody does technical SEO properly. That is not what the data says. 87 percent of the product pages we could measure carry valid Product schema, and 83 percent carry Offer markup with price and availability. These are competent teams with real technical capability.
The gap is narrower and stranger than incompetence. These brands have carefully described what they sell. They have almost entirely failed to describe the evidence that it works.
Only 2 of the 23 brands we could measure, 9 percent, publish machine-readable Review markup. Yet nearly every one of these sites displays customer reviews on the page in a format a human can read and a machine cannot reliably parse.
That distinction matters more in AI search than it ever did in traditional search. A blue-link result rewards a page for being relevant. An AI answer engine is doing something harder: it is assembling a recommendation and it has to justify the pick. Ratings, review counts, and structured evidence are the raw material for that justification.
Schema.org coverage on product detail pages
Share of the 23 brands whose product page we could retrieve and parse. Collected August 17, 2026.
The top three bars describe the product as an object: what it is, what it costs, where it sits. The bottom three describe it as a choice a person made and rated. The industry has built the first half of the picture and left the second half in HTML that only a human can read.
We read every robots.txt file we could reach, 28 of 30, and checked them against twelve named crawlers that feed ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews.
Exactly 0 of 28 brands named a single one of those crawlers in any directive. Not to allow them. Not to block them. The names simply do not appear.
We expected to find an industry quietly locking AI engines out and damaging itself by accident. That is not what happened. Only 1 brand in the sample blocks AI crawlers at all, and it does so through a blanket rule rather than a considered choice. The rest have left the door open without noticing there is a door.
7 brands returned 403, 429, 404, or a failed connection to ordinary automated requests. It is tempting to report that as evidence they block AI crawlers. We are not doing that, because real crawlers such as GPTBot and Googlebot authenticate by published IP range rather than by user-agent string, and a bot manager rejecting an unverified request tells you nothing reliable about how it treats a verified one. Those brands were excluded from every percentage in this study rather than counted as failures.
Two brands in the sample, Bulletproof, Ka'Chava, carry the full stack on their product pages: Product, AggregateRating, and Review together. That combination gives an answer engine a complete, machine-readable case: here is the product, here is what it costs, here is how it is rated, and here is what buyers said.
At the other end, 9 brands declare Product schema with no rating and no review markup whatsoever. They have told the machines what they sell and nothing about whether it is any good.
3 brands, HUM Nutrition, Legion Athletics, Thorne, run product detail pages with no Product schema at all, which is the most straightforward gap in the study and also the easiest afternoon of work on this list. One brand, Thorne, ships structured data on its homepage that fails to parse as valid JSON, which means it is doing the work and getting no credit for it.
We named the brands doing this well and the specific, factual technical gaps we measured. We did not rank brands worst-to-best or imply anything about product quality, because this study measured markup on a single day and nothing else. A missing schema block is a two-hour fix, not a verdict on a company.
Sample. 30 direct-to-consumer supplement and nutraceutical brands, drawn from brands named across independent 2026 "top DTC supplement brand" listings plus large well-known DTC supplement retailers. This is a convenience sample of prominent brands, not a random sample of the industry, and it skews toward well-funded companies. If anything, that biases the results optimistic.
Collection date. August 17, 2026. Sites change. Any brand here can close these gaps in an afternoon, and we hope some do.
robots.txt. Retrieved from the root of each domain and parsed for directives naming twelve crawlers: GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, ClaudeBot, anthropic-ai, Google-Extended, CCBot, Applebot-Extended, Bytespider, and meta-externalagent.
Structured data. Every application/ld+json block on the homepage and on one real product detail page per brand was parsed and every schema.org @type recorded. Product pages were discovered by following product links from the homepage, so the page measured is one a visitor could actually reach.
Exclusions. 7 brands could not be measured and were removed from all denominators: Ritual, Seed, AG1, Your Super, Garden of Life, Care/of, MyProtein. Percentages are over the 23 brands with a retrievable product page.
What we refused to measure. We did not test whether any AI engine actually cites these brands, because citation results vary by prompt, user, region, and day, and a single set of queries would not survive scrutiny. We did not evaluate product quality, ingredient claims, design, or copy. We measured only what is machine-checkable and reproducible.
If you run a supplement brand and you want to be citable by AI search, the order of operations is unusually clear, because most of the expensive work is already done.
You almost certainly collect reviews. Most brands render them in HTML that reads fine to a person and means nothing to a parser. Adding AggregateRating and Review markup to product pages is the single highest-leverage change on this list, and it is the one the fewest brands have made.
One brand in this sample publishes structured data that does not parse. Broken markup is worse than no markup, because you believe you are covered. Run your product pages through a validator and fix the parse errors before writing any new schema.
35 percent of measurable product pages carry FAQPage markup. Note that Google deprecated FAQ rich results in May 2026, so this will not win you a SERP dropdown anymore. It still gives an answer engine clean question-and-answer pairs to work from, and Google has confirmed it continues to read the markup to understand pages.
Whatever you choose, choose it. Right now the overwhelming majority of this industry has no position in robots.txt at all. If you want AI engines quoting your product pages, confirm nothing at the CDN or WAF layer is quietly refusing them, because that decision often lives with an infrastructure vendor rather than with marketing.
None of this makes an engine recommend you. It makes you legible enough to be recommended. Legibility is necessary and not sufficient, and most of this industry has not cleared the necessary part.
Two things, both objectively verifiable and reproducible. First, whether each brand's robots.txt file allows or blocks the named crawlers that feed ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. Second, which schema.org structured data types appear on the brand's homepage and on a real product detail page. We did not measure design, copy quality, product efficacy, or actual citation rates inside any AI engine.
Answer engines have to decide what a page says without a human reading it. Structured data is the machine-readable layer that states, unambiguously, that this is a product, this is its price, this is its rating, and these are the questions it answers. Without it, an engine has to infer all of that from prose, and inference loses to explicit markup when the engine is choosing which of ten brands to name.
In this sample, yes. 87 percent of measurable product pages declare Product schema, so these brands clearly have technical SEO capability. But only 9 percent carry Review markup and only 48 percent carry AggregateRating. When someone asks an AI engine for the best supplement for a given goal, the engine is assembling a recommendation, and recommendations lean on evidence that other people rated the thing well.
No, and anyone promising that is overselling. Structured data determines whether an engine can cleanly use your page once it has found it. It does not determine whether the engine finds you, which depends on your broader authority and how often other credible sources mention you. Markup removes a barrier. It does not manufacture demand.
7 of 30 brands returned an error to ordinary automated requests: 403 Forbidden, 429 Too Many Requests, a 404 on the homepage, or a failed connection. That is normal bot-management behavior at the CDN layer and we are explicitly not claiming it blocks real AI crawlers, because verified crawlers authenticate by IP range rather than by user-agent string. We simply could not measure those sites, and we excluded them from every percentage rather than guessing.
Essentially none. Across the 28 brands whose robots.txt we could read, 0 explicitly named a single AI crawler in any directive, allow or disallow. The industry has not made a deliberate decision about AI access. It has made no decision at all, which is itself the finding.
Yes, and that is the point. Every measurement comes from a public robots.txt file or the public HTML of a homepage and a product page. The methodology section states the sample, the date, the crawler list, and the exclusions. Run the same checks and you should get the same answers, allowing for sites changing after our collection date.
PELORA Marketing is a marketing agency in Newport Beach, California that works with supplement, nutraceutical, and regulated health brands nationwide. We ran this because we kept telling clients that AI visibility is a structural problem rather than a content problem, and we would rather publish the measurement than keep asserting it.
Supplements were one group in a larger benchmark. We scored 121 businesses across med spas, weight loss clinics, men's health clinics, marketing agencies and DTC supplement brands using a single 100-point rubric. Supplement brands finished last. Read the Local AI Findability Score.
We run the same audit on your site, free, and send back exactly what is missing and what it would take to fix. No obligation, and we will tell you if the answer is that you do not need us.