Newsletter · · Ashutosh Agarwal

Advertising to AI Agents Just Got Real - Digital Ads and Retail Media Weekly - Week of August 30, 2026

Digital Ads & Retail Media Weekly for the week of August 30, 2026. Podcast synthesis on Time Magazine selling sponsored answers into the machine-readable web, Perplexity blocking them and threatening trust-score downgrades, a fake deodorant brand built for eleven dollars that ChatGPT recommended first, and the state attorneys general seeking limits on Meta's algorithmic feed.

Digital Ads & Retail Media Weekly

Week of August 30, 2026: Advertising to AI Agents Just Got Real

Here's a sentence that would have sounded like science fiction a year ago: this week, the most interesting story in advertising wasn't about reaching people at all. It was about reaching the robots that increasingly shop for people.

TL;DR

  • A new ad channel is being born, advertising to AI agents, not humans. Time Magazine started slipping sponsored answers into the machine-readable version of its site that AI crawlers read, and Perplexity promptly blocked it and threatened to lower publishers' trust scores. The fight over who controls the "AI shelf" has officially begun.
  • You can game ChatGPT's recommendations for about $11. A researcher invented a fake deodorant brand, spent an hour and eleven dollars building a thin website around it, and three weeks later ChatGPT was recommending it first in four out of four answers. Discovery is moving to a channel brands don't control and can't yet measure.
  • Meta's real ad risk this week was legal, not competitive. A California jury and judge handed down a public-nuisance finding, and 29 state attorneys general are seeking up to $200 billion in damages plus limits on the algorithmic feed, the exact machinery that makes the ads work.

What's new

1. Advertising is quietly forking into two internets, one for humans, one for machines. On The Marketing AI SparkCast with Aby Varma (Aug 23), host Aby Varma (who runs a marketing-AI advisory called Spark Novus) and co-host Matt Sear walked through a story first reported by Digiday. Time Magazine began publishing stripped-down, machine-readable versions of its pages for AI systems, then, working with an ad-tech company called Mobian, turned those pages into advertising inventory. Sponsored, FAQ-style content from brands like Ally Bank and the Project Management Institute was inserted into the version of the page served to AI crawlers, not the version humans see. The goal, as Varma put it, was to feed AI systems "current brand-approved information that could potentially influence how advertisers appear in future AI answers."

Why it matters: this is a preview of a brand-new ad format that sits underneath the answer an AI gives you. If it works, it's a threat to the open-web ad model as we know it and a land grab for whoever sets the rules. And that fight started immediately, see the next item.

2. Perplexity fired the first shot, and it tells you who holds the power. In the same episode, the hosts noted that in early August, Perplexity confirmed it was blocking Time's marked-down ads from influencing its agents and its user-facing results. It called the format "deceptive" (the ads showed up in the crawler-facing page but not the human one) and warned that publishers trying this "could receive a reputational downgrade," a lower trust score in Perplexity's search index. As of August 21, the two sides were reportedly "working it out."

Why it matters: Varma's framing was that the AI platforms "are becoming the gatekeepers." Whoever runs the AI answer box gets to decide what counts as an ad, what counts as deception, and which publishers get demoted. That's enormous power, and it's being defined right now with no agreed industry standard for disclosure or measurement. His blunt advice to marketers: "AI agents may be the next media audience… but right now, nobody knows the rules and nobody can really prove the ROI." Follow it, demand transparency, don't call it a proven channel yet.

3. A fake brand and $11 beat every real advertiser in its category on ChatGPT. The most memorable story of the week, also from the SparkCast, was co-host Matt Sear's account of a researcher named Deanna Burke. She spent "about $11 in about one hour" inventing a natural-deodorant brand called Morrowind: a three-page website, some formulation copy stuffed with the right keywords, an AI-generated image, and a Substack essay. No reviews. No ad spend. The product doesn't exist. Three weeks later, ChatGPT "was recommending Morrowind first in four of four answers" for people searching for a niche type of deodorant (the "magnesium and baking soda, irritation-focused" crowd).

Sear's takeaway is the line to remember: "The AI shelf that's feeding these recommendations is cheap and easy to game right now… Discovery is moving to a channel that you don't control. And it can be manipulated by anyone with a domain, an hour, and very, very little money."

Why it matters: for now, this only works on narrow, long-tail questions where real competition is thin. As Sear said, it "does not extend to Dove brand deodorant." But if consumers keep treating AI recommendations as pre-vetted (and the hosts think they increasingly do), then getting mentioned in the AI answer becomes the new shelf placement, and it can be bought or gamed far more cheaply than a search ad or a retail-media slot today.

4. Meta's biggest ad-thesis risk this week came from a courtroom. The one place any of the big ad platforms showed up was Telltales (Aug 23), and a caveat matters here: Telltales is an AI-generated markets-recap show (it says so itself: "produced entirely with AI tools, and both voices you're hearing are AI generated"). So treat it as a tidy summary of publicly reported facts, not as expert or insider commentary. With that flag raised, the facts it recapped on Meta are real and they matter for the ad thesis:

  • A California jury found Meta liable on 75,000 violations of state consumer-protection law, and a judge separately found its platforms constitute a "public nuisance." The settled part is $375 million in penalties.
  • The unsettled part is the scary one: 29 state attorneys general are in federal trial alleging Meta deliberately designed Facebook and Instagram to be addictive, seeking up to $200 billion in damages plus limits on the algorithmic feed. A former Meta engineering director testified that leadership treated child safety as secondary to growth.

Why it matters: the dollar figure gets the headline, but the real risk to advertisers is that last phrase, limits on the algorithmic feed. The feed's engagement loop is the engine that sells the ads. The show's sharpest point was that a public-nuisance finding becomes "a template every other state attorney general can now run." Meanwhile Meta guided Q3 revenue to $61 to $64 billion (midpoint below where the Street sat) and is cutting staff across WhatsApp, Instagram, and Reality Labs for the second time this year.

The debate

Bull case: AI is expanding the pie and the walled gardens win again. Every one of this week's stories is, at bottom, evidence that a new advertising surface is being created: the AI answer. New surface means new inventory, new budgets, new formats. And the companies with the models, the data, and the distribution to own that surface are the same giants who already own search and social. When a fresh channel appears, the walled gardens usually end up collecting the toll. The Perplexity episode is Exhibit A: the platform, not the publisher, decides what runs.

Bear case: the ground is shifting under everyone who sells ads today. Look at the same stories from the other side. The $11 fake-brand experiment says the value of the paid placements advertisers buy today can be undercut by anyone who learns to game the AI answer. The Time and Perplexity standoff says the open web's publishers are being turned into raw material for someone else's answer box, with their monetization at the mercy of a "trust score" they don't control. The Duke CMO survey the SparkCast cited adds a sobering budget backdrop: AI use in marketing has more than doubled in two years, yet training budgets fell to just 3.8% of marketing spend and marketing headcount growth dropped more than half, so a lot of this "AI advertising future" is being bought without the people or process to make it pay. And Meta's week is a reminder that the most reliable ad machine on earth still carries a tail risk: regulators reaching in and throttling the feed itself.

The swing factor: whether AI-answer advertising becomes a real, measurable, disclosed channel, or stays a Wild West that erodes trust faster than it creates budget. Right now, nobody can prove the ROI. That's the whole ballgame.

Stocks in play

Meta (META)

  • Bull: Still the best-performing ad engine in the business, and best-positioned to own AI-driven ad formats across Facebook and Instagram. The stock has already de-rated hard (Telltales pegged it down 32%), so a lot of bad news is arguably in the price.
  • Bear: The legal overhang is now open-ended, up to $200 billion sought and, more dangerously, potential limits on the algorithmic feed that powers ad targeting. Q3 revenue was guided below the Street, and a second round of layoffs this year signals the company is bracing, not sprinting.
  • Next number to watch: Which state attorney general files next, and whether it lands before or after the next capital-spending update, plus whether Q3 revenue comes in at or above the $61 to $64B guide.

Walmart (WMT) (read-through only)

  • Bull: Traffic is the lifeblood of a retail-media network, and Walmart is spending a $2.9 billion tariff refund on price cuts to buy exactly that. More trips means more valuable ad inventory for Walmart Connect down the line.
  • Bear: U.S. comparable-sales growth was the slowest in six years, and the guidance raise leans on a one-time customs windfall. As the Telltales host put it, "a customs windfall spent on buying traffic… works right until the refund stops arriving."
  • Next number to watch: Whether comps re-accelerate on their own once the refund-funded price cuts lap, and any fresh disclosure on Walmart Connect's growth rate.

The platforms exposed to AI-answer advertising (GOOGL, and privately, OpenAI and Perplexity)

  • Bull: Whoever runs the dominant AI answer box inherits the most valuable ad real estate of the next decade and gets to write the rules. Google and the large AI platforms are the natural landlords.
  • Bear: If the answer box gets gamed or loses user trust (the $11-brand problem), the value of that real estate collapses before it's monetized. And search advertising, Google's crown jewel, is precisely what an AI answer that needs no clicks threatens to disintermediate.
  • Next catalyst to watch: Any move toward an actual, disclosed ad format inside ChatGPT, Gemini, or Perplexity, and how the Time and Perplexity dispute resolves, since it's the first real test case for the rules.

Read-throughs

  • Ad measurement and verification (DV, IAS, RAMP, CRTO, MGNI, PUBM): The week's big theme, measuring and verifying ads inside AI answers, is precisely the problem these companies exist to solve. If AI-answer advertising becomes real, this is the group with the most to gain from a new "how do we even measure this" scramble. Worth watching as a second-derivative play.
  • Retail-media operators (WMT, CART): Walmart surfaced only through its comp-sales print, not through the ad business itself, so the roughly $129B retail-media theme is still waiting on a fresh operator read.

What changed vs. last week

The conversation's center of gravity moved. A week ago the sector talk was about the usual battlegrounds, CPMs, retail media, CTV. This week the fresh thinking was about advertising to machines and gaming AI answers. That's either a one-week blip or the first sign that the next front in digital advertising opens inside the AI answer box. We'll know more as the platforms report and the ad-tech shows catch up to it.