Every CMO watching search traffic erode has heard some version of the same pitch: use AI to make content faster, cheaper, more personalized. Tory Brangham, chief commerce officer at People Inc., did something a board would actually ask follow-up questions about. She looked at a decade of photography her editorial teams had already shot, for magazines like People and Better Homes and Gardens, content nobody had ever priced as a revenue asset, and used computer vision to turn it into a shopping engine now running 29 million transactions a year. No new campaign. No new budget line for content production. The growth came entirely from an archive the company already owned.
An archive most finance teams would call a sunk cost
Brangham's underlying thesis predates the AI framing by years and holds up as a pure piece of marketing logic. The instant a reader looks at a photo and thinks that dress is nice, or I want that couch, is the single highest-intent moment inside the entire piece of content. Almost every publisher wastes it, chasing the reader afterward with an affiliate link at the bottom of the post, once the impulse has already cooled. Her read was that the interruption itself, forcing someone to leave the article to go find the product, is what kills the sale. It's the kind of insight that sounds obvious once stated and expensive to have missed for twenty years.
The system built with Shopsense AI removes that interruption. Computer vision scans images and video already live across People's properties, identifies the actual products inside them, and turns them into shoppable links without an editor tagging a single item. The reader never leaves the page. What used to require a human noticing and linking a product, one article at a time, now runs automatically across an entire archive.
The traffic problem this was actually built to solve
People Inc. is the company formerly known as Dotdash Meredith, formed in December 2021 when Dotdash bought Meredith's magazine assets, including what remained of the old Time Inc. titles, for 2.7 billion dollars. It took the People Inc. name on July 31, 2025, in the middle of absorbing the same traffic decline every publisher with a large library of how-to and buying-guide content is now facing, as Google reroutes search traffic through AI Overviews and AI Mode instead of a page of blue links. AdExchanger has covered the company looking inward for growth specifically because its search-referred traffic keeps shrinking. In the company's own Q2 2025 results, performance marketing revenue rose 14 percent and affiliate commerce specifically grew 25 percent, at the exact moment traffic feeding the rest of the business was declining. For a board asking where growth comes from when a traffic channel is dying, this is one of the more concrete answers on record: an owned asset that Google's algorithm cannot touch.
The math that actually explains why finance signed off
The part of this story most coverage skips is the cost structure, and it's the part that should matter most to anyone who has to defend an AI budget line. Shoppable video, where most of the industry's current investment is going, requires new production for every unit of commerce content: a shoot, a host, editing. Shopsense's model runs entirely on content that already exists, so the marginal cost of tagging a photo that already has traffic against it is close to zero. Every dollar People Inc. spent here went toward unlocking value sitting inside content it had already paid to produce years earlier, for reasons that had nothing to do with commerce. That distinction, monetizing an existing asset versus funding new production, is the difference between an AI expense and an AI return.
The numbers a board would actually ask to see
Adweek's June 2026 Commerce All-Stars feature put real figures behind the claim. Clickthrough rates on the AI-identified product placements ran up to 100 times higher than standard IAB display ad benchmarks. Earnings per click rose by up to 70 percent. The program drives more than 29 million transactions a year. Brangham's own team has grown from 50 people to more than 300 since she joined the company in 2017, the kind of headcount growth a finance department only approves around a strategy that is demonstrably working.
Worth stating plainly: these are company-reported figures, run through Adweek's editorial process rather than an independent audit, which is standard for coverage like this. What lends it more weight than the average AI success story is that an outside trade publication chose to name Brangham specifically for the results, rather than the numbers living only on a company website, and a second, entirely separate data point corroborates it from the technology side.
The one benchmark most AI vendors never publish
That second data point landed on June 2, 2026, when Shopsense AI announced the Shoppable Intelligence Model, built specifically to identify products in content rather than adapted from a general-purpose vision system. The company benchmarked it directly against OpenAI's CLIP and Google's SigLIP2, two of the most cited computer vision models in the field, on public fashion datasets researchers use to evaluate exactly this kind of work. The published results show Shopsense beating CLIP by up to 77 percent on image-to-image retrieval accuracy and up to 74 percent on text-to-text accuracy, and beating SigLIP2 by 34 to 60 percent across retrieval tasks. Any skeptical marketing leader can pull the same public benchmarks and check it independently.
Here is the number that should matter most to anyone evaluating an AI investment: a 10 percentage point gain in product-identification accuracy produced a 24.5 percent increase in shopper clickthrough rate. A technical metric tied to a stated business outcome, published where it can be verified, is rare in AI marketing coverage. Most claims arrive as an unquantified productivity story or a viral demo with no revenue figure attached anywhere near it.
What competitors could copy tomorrow, and what they couldn't
The question worth asking before treating this as a durable advantage: how much of it is actually defensible. Shopsense's accuracy lead over CLIP and SigLIP2 will likely compress as competing tools mature over the next few years, so the technology itself is not the moat. What's harder to replicate is the archive and the audience relationship built around it, years of published photography and reader trust that a competitor cannot acquire by simply licensing similar software. Shopsense supplied the mechanism. People Inc. supplied the years of content that made the mechanism worth anything. That distinction matters for any leadership team evaluating whether an AI vendor is selling a genuine edge or a commodity that every competitor will have within eighteen months.
A three-founder company outperforming Google and OpenAI's benchmarks
Worth pausing on the scale of the company behind this. Shopsense AI was founded in 2023 in Redwood City by Eve Friedman, Bryan Quinn, and Glenn Fishback, on a single seed round of 2.2 million dollars, according to public funding records. That is an unusually small balance sheet to be running commerce infrastructure for one of the largest magazine publishers in the country, let alone outperforming benchmarks from two of the best-funded AI labs in the world. Since launching in April, Shopsense has added Paramount, Univision, Tastemade, and Nexstar Media Group's The CW as clients, and in December 2024 partnered with Bell Media to bring shoppable placements to Canadian shows on CTV. None of these media companies has a reason to keep paying for something that isn't moving revenue, which makes the client list itself a form of evidence.
The bottleneck affiliate commerce never actually solved
Content commerce is not a new category. Publishers have chased it for close to two decades, starting with coupon sites and cash-back plugins, turning it into serious revenue for the better operators over roughly the last ten years. Forbes reported the New York Times' Wirecutter drove more than a billion dollars in gross merchandise value in both 2023 and 2024, built almost entirely on affiliate commerce before it added advertising for the first time last year. Bustle Digital Group saw an earlier version of this during a pandemic-era quarter of 2020, when its affiliate business grew 84 percent year over year while advertising revenue fell by roughly a third in the same stretch. Industry-wide, affiliate marketing spend hit an estimated 15.7 billion dollars in 2024, and Rakuten research from 2022 found 84 percent of publishers already running some kind of affiliate program. The bottleneck was never the strategy, it was always the labor: an editor manually placing links one product at a time, which capped how much of any archive could ever get monetized.
That labor constraint is what computer vision actually removes. No editorial team has the hours to go back through a decade of archived photography and tag every product in it by hand. A model does, and it tags new content the moment it publishes rather than months later. The 29 million transactions a year mostly demonstrate what happens once a strategy publishers already believed in stops being limited by headcount.
Why the industry's money is chasing the harder version of this
Most current industry investment sits in shoppable video. Firework, which counts Bloomingdale's and Hugo Boss as clients, along with Tolstoy and Bambuser, ask brands to produce new interactive video built specifically to be shopped. The visual commerce platform category overall is projected to grow from 2.5 billion dollars in 2024 to 6.8 billion dollars by 2033, and most of that growth assumes brands keep funding new video production indefinitely.
Shopsense avoids that production cost entirely, and it's the detail that should matter most to any marketing leader comparing the two approaches. A publisher shooting a home tour or a recipe video was always going to make that content anyway, for editorial reasons unrelated to commerce. Shopsense finds the transaction value sitting inside it afterward, at close to zero marginal cost. Between an approach that requires funding new production indefinitely and one that monetizes what already exists, only one of them survives a budget cut.
Where this same logic applies outside publishing
The underlying mechanism, computer vision finding transactable value inside content that already exists, extends well past magazine publishing. A hospitality brand with years of guest photos and property tours is sitting on a comparable archive, every piece of furniture and fixture in those photos is a potential product link nobody has connected. A real estate brokerage's listing photography carries the same latent value, appliances and staged furniture currently monetized only through the sale of the property itself. A food media brand's recipe library is full of cookware and ingredients that could be tagged the same way People Inc. tags a couch. None of it requires producing anything new. It requires recognizing that a content archive is a balance sheet asset, not just a marketing cost.
The standard this sets for the next AI pitch on your desk
There's a broader lesson here that has nothing to do with commerce specifically. Most AI claims that reach a CMO's desk arrive as an unquantified productivity story with no way to check it. Shopsense's numbers are different because they trace to something concrete: a named benchmark, a named comparison, a technical metric tied to a business outcome. The ten-point precision gain tied to a 24.5 percent clickthrough lift is the kind of specificity that should be the baseline for evaluating any AI vendor claim, not the exception. A two-year-old company running on 2.2 million dollars produced that level of proof while landing five major media partners inside two years, worth remembering the next time a bigger name or a bigger balance sheet gets mistaken for evidence that something actually works.
