[Your Brand] on the AI shelf

Assistants opened 3 pages on your-brand.example across 20 answers and named [Your Brand] in none of them.

• https://your-brand.example/wellness/supplements/types-of-magnesium - read in 10 answers, named in 0
• https://your-brand.example/wellness/supplements/magnesium-glycinate-vs-citrate - read in 5 answers, named in 0
• https://your-brand.example/wellness/supplements/magnesium-heart-health - read in 5 answers, named in 0

The page was read and quoted, so it is reachable, indexed and relevant to the question - it cleared every bar but the last one. What the assistant could not lift out of the text is the claim "this is [Your Brand], and this is what it does". That is what happens when a page speaks as "we" and the name appears only in the header, the logo and the tab title.

• Name [Your Brand] in the body text, not only in the header, the logo and the tab title
• Give the page a sentence an assistant can quote as-is: "[Your Brand] is X, suited to Y"
• Check the name is served as text rather than painted in by a script or baked into an image
• Add Organization/Product markup carrying the name - to an assistant that is a direct answer to "whose page is this"

These pages already clear the bar - they are read in 20 answers of the measurement. Nothing here has to be negotiated with anyone: it is the only work in this report that is entirely on your side.

In answers where [Your Brand] ranks below a competitor, the higher-ranked entrant is credited with these properties:

• Third-party tested - 38%

These are the specific properties assistants cite when they justify a pick. They are visible, checkable, and mostly fixable.

• Where [Your Brand] lacks a property, treat it as a roadmap item and price the gap
• Where [Your Brand] has one but is not credited for it, the property is not readable - put it in page text and structured product data, not only on packaging
• Third-party testing and certification recur across this category; if [Your Brand] holds either, make it impossible to miss

The most direct lever in this report - it is the difference between being listed and being recommended.

[Your Brand] appears in 6% of answers but is named first in only 0%. When it is named, it sits on average at place 4.9 in the assistant's list of brands. These are the questions where it comes closest, with the brand that takes first place in those same answers.

• "practitioner grade magnesium brands" - named in 6 of 15 answers, best place 3, first place goes to Thorne
• "best magnesium supplement" - named in 5 of 15 answers, best place 3, first place goes to Pure Encapsulations
• "best magnesium brands 2026" - named in 6 of 15 answers, best place 2, first place goes to Thorne

Being named is a coverage problem; being named first is a positioning problem. Assistants lead with whichever brand most cleanly matches the intent in the question - "best overall", "best value", "best for women". [Your Brand] currently reads as a safe second choice rather than as the answer to a specific one.

• Start at the top of that list: it is the question where [Your Brand] is closest to first place, and it has already reached place 3 there
• Read what Thorne publishes for that question - it takes first place in the same answers that name [Your Brand]
• Answer that question in the first line of the page, in the words the question uses, not in a spec table below the fold
• Keep the claim checkable - a certification, a test result, a number: assistants lead with the brand whose claim they can attribute to something

[Your Brand] is effectively never the first name in the answer today, so the entire first-mention share is still open. The first three places are where an assistant's recommendation actually lands - the rest of the list is read as also-rans.

Assistants opened 3497 links to answer the Magnesium questions in this measurement, and they took them from a narrow set of domains: healthline.com alone accounts for 7% of them. [Your Brand] is named in answers that read 4 of the 5 domains below, and never in the answers that read the other 1.

• healthline.com - 7% of links, read in 175 answers, named in 17
• amazon.com - 4% of links, read in 65 answers, named in 10
• pmc.ncbi.nlm.nih.gov - 4% of links, read in 76 answers, named in 0
• mcpress.mayoclinic.org - 2% of links, read in 86 answers, named in 6
• mitohealth.com - 2% of links, read in 74 answers, named in 16

Grounded assistants cite before they recommend. A brand that is not on those pages is usually not in the answer. The last number on each line is how often reading that domain ended in an answer that named you - a domain read this often with a low number beside it is a page the assistants already trust and you are missing from.

• Work the list from the top down - most-read first, and among those the ones where [Your Brand] is named least
• Submit products to the testing and review outlets in that list - they publish the pages assistants read
• Audit any listing [Your Brand] controls outright: it is the one page here you can change today

The slowest of these moves and the most durable. A placement lands in front of every answer that already opens that domain - healthline.com alone is read in 175 answers of this measurement. Citation placement compounds: it feeds every future answer.