the people behind ai’s recommendations

Across the portfolio, 25.9% of what AI cites is a page with a named author behind it. Video creators, grouped with community and social above, add another two points. A further 3.8% is published by an organization’s name rather than a person’s. The rest, close to six in ten citations, is unattributed. This includes brand sites, retailers, directories, and publications that do not credit their writers on the page.

By surfacing the individuals that drive the quarter of all AI citations, we give brands a clear path to act upon. A human can be pitched, sent product, invited to visit, or briefed. A retailer’s product page has a more challenging path to inclusion.

This representation varies across models. Grok lands on a named author most often and ChatGPT least, which fits what Report No. 02 found about ChatGPT’s Wikipedia habit.

On a typical brand, the Centium dashboard names 74 people. The single most cited writer holds a median 10.6% of everything AI reads from people in that brand’s dataset. The top three hold 23.8%. The top ten hold 51.2%.

The #1 voice is usually one heavily cited piece plus a body of work. Their single best article carries a median 61% of their citations, but rarely did they publish just one article. Only one in four of the top writers (63 of 269 brands) had just one cited article.

The ten most-read writers on a brand carry half of what AI reads from people about it. We compared them with everyone ranked below.

Three things separate them, and one thing does not.

Depth over breadth. A top voice is someone AI reaches for on several pieces, each pulled many times: two or more cited articles for 52% of them, against 13% in the tail, and 10.8 citations per article against 2. The tail is one piece, pulled twice.

Writers are most impactful. Video creators are a quarter of the long tail and 4% of the top ten, typically driving less citation per piece. A writer accumulates across a body of work.

One outlet. 94% of top voices publish on a single outlet in the data. The freelancer spread across titles is rare.

Brands have room to improve. 36% of the top ten authors mentioned the measured brand at least once in their cited work. On 147 of 269 brands, the single most read writer in the category has never named the brand.

Half of everything AI cites from named writers is a roundup, and among the top ten voices it is 57%. Add hands-on reviews and it is two in three of their cited pieces. The pieces AI keeps returning to sit on evergreen slugs like /best-gravel-bikes and get updated in place, which is also why so many of them are years old.

What that looks like, from some of the most cited pieces of the sector leaders:

• “Best gravel bikes”, Cyclingnews, first published 2022, cited 191 times
• “Best base layer”, OutdoorGearLab, 2026, cited 218 times
• “Best lesser-known resorts for experts”, SKI, 2025, cited 111 times
• “Best sugar-free gum”, Sporked, 2025, cited 126 times
• “Ikon Pass vs Epic Pass”, OnTheSnow, 2026, cited 60 times
• “Family-friendly cities in the USA”, Family Can Travel, 2025, cited 29 times

The mix shifts by sector. Snow sports and apparel are almost entirely roundups and reviews, and the pieces are fresh: the median cited article is under a year old. Cycling leans on old pieces (median 1.6 years) written by people who have mostly already named the brand. Travel and hospitality are guides and itineraries. B2B has no editorial genre at all, for a reason covered next.

We classified every outlet by how many different writers the portfolio has seen on it: one or two means an independent or personal site, five or more means a publication with a staff.

In gear, AI reads the testers at the publications: OutdoorGearLab, Cyclingnews, BikeRadar, Bicycling, Switchback Travel, SKI. The top three voices on a cycling brand are staff testers 94% of the time.

In travel, AI reads independent bloggers on their own domains. The writers who appear on the most destination dashboards run sites like Family Can Travel, Travel Babbo, The Big Outside, Hike Bike Travel and Explore With Alec. None of them work for a publication. For a destination or a resort, the person to host is not the magazine’s travel editor; it is the blogger whose national-parks itinerary AI keeps reaching for.

In B2B, “independent” mostly means programmatic or vendor-owned. The names that recur across software and services dashboards belong to templated statistics and directory sites that stamp the same author on thousands of “best X in city” pages, and to the content teams at software vendors like SafetyCulture and monday.com. AI reads them anyway. For a B2B firm, the most influential voice in the category is often not a journalist, and the earned-media play is to become the cited page rather than to pitch the writer.

The writer whose work AI cited most in each sector. Citations are counted across every brand we track in the sector; articles are the distinct pieces AI reached for; coverage is the share of the sector’s brands on whose dashboard the writer appears. Ski resorts and ski destinations are one sector: they read the same resort guides whether the brand is a mountain or the town below it. One name per sector here. Centium subscribers see these names, the dozens behind them in their own category, and which of their articles named the brand, on the Sources tab of their dashboard.

Two things stand out. The first is that some sectors have a bench and some do not. Tara Seplavy’s Bicycling roundups were cited 1,461 times across the cycling brands we track, and she appears on 97% of them. Ski is more concentrated still: Samantha Berman’s SKI resort guides appear on 96% of the ski resorts and destinations we track.

Travel has a bench too, but it is regional. The writer AI reads about a destination is the one who covers that place: Leigh McAdam’s Canadian Rockies guides carry 177 of her 222 citations on a single British Columbia destination, and a San Diego blogger leads a San Diego hotel. Those writers rarely recur beyond their region, and the national roundup writers who do recur appear on a dozen destinations for a handful of citations each. A cycling brand or a ski resort competes for the attention of a small group of writers everyone already knows. A destination’s list is local.

B2B is the extreme case and is left off the table. AI reads named people there as often as it does in travel, but every category reads different people, so no name accumulates across the sector’s brands. For a B2B firm the pitch list exists only per brand.

The other is how far the same people travel. 3,097 of the 13,488 writers appear on more than one dashboard, and the gear testers span sectors: the same Bicycling and Cyclingnews staff surface on apparel, nutrition and software brands, because a roundup of cycling clothing or training apps is still cycling journalism. Pulled together across every sector, five writers have passed a thousand citations across the brands Centium tracks. They are the people AI reads most when it recommends those brands.

The pitch list matters because of what the other side of it is made of. We counted, for every brand, how many of its ten most-read voices have named it, and set that against the brand’s AI visibility rate: the share of category-search answers that mention it.

A brand none of its top ten voices has named is mentioned in 10% of AI answers. A brand all ten have named is mentioned in 78%, and the rate climbs with every writer in between, 5.75 points of AI visibility per voice. It holds inside every sector, so it is not gear brands lifting the average, and it holds after controlling for citation volume, so it is not big brands being big.

AI recommends the brands it finds in the pages it reads, so a brand named by more of its top voices will always be mentioned more often. The chart does not prove that a single pitch raises visibility. What it shows is how large the effect is, and that it is consistent in every industry we track. That is why targeting the writers who have not named you yet is one of the most effective strategies for growth.