AI Recommendation Engines Compress Retail Product Visibility for Brands

Full analysis

What's actually being decided: whether operators should redirect budget toward AI-visibility work (GEO, owned-media content) now, or wait until AI-driven purchases are a real share of the funnel. This is easy to undo. Nobody signs a multi-year GEO contract; you can start and stop this quarter. That argues for cheap tests, not a strategy offsite. What sets the clock is the H2 budget cycle, when GEO line items hit their first CFO review.

The Market Analyst follows where budget moves when organic discovery gets compressed. If AI shows a narrow shelf, the discovery you used to earn for free now costs money, and that money flows to placements that guarantee shelf presence. Retail media is the clean winner: Walmart Connect, Amazon Ads, Kroger, Roundel all sell guaranteed visibility that AI can't randomize away. The loser is the comparison-shopping and mid-tier aggregator layer, because a model summarizing options doesn't need a middleman that also summarizes options. For a non-specialist: when the free path to shoppers narrows, brands pay the retailer to stay visible, and the retailer wins.

The Skeptic sees the GEO panic as a consulting invoice looking for a strategy memo. For this to matter at scale, AI answers have to drive a real slice of purchases, and they don't yet. Most transactions still run through Amazon, retailer apps, and Google Shopping. Droesch questions whether moving from 18% to 20% AI visibility justifies the spend. When the analyst raising the alarm can't make the ROI case himself, the story is anxiety, not economics. Nike and Adidas dominating broad queries is brand equity doing exactly what brand equity does. In plain terms: the scary shelf is real but tiny, and small does not justify a new budget line.

The Operator sees two mandates hit the same budget cycle and collide. Someone buys a GEO package to lift AI visibility. Someone else runs the brand.com content team that Droesch's own data says is the actual lever, and that team is under-resourced. The GEO line breaks first, cut at the quarterly review for being unattributable. The multi-condition query problem breaks second: you can engineer "best running shoe" but not "best waterproof trail shoe under $90 for flat feet," and no vendor can promise you that slot. Retail media reps should be dialing brand partners this quarter with exactly this pitch. In plain terms: pay for placements you can guarantee, not for a lottery ticket in the model's answer.

The CFO approves owned-media content and declines GEO, and here's the split. Brand.com content is an asset I keep: it feeds AI citations, it feeds search, it feeds the site itself, and I own it. GEO is a fee I pay a vendor to move a number I can't attribute from 18 to 20. One compounds, one evaporates. Show me an incrementality test that isolates AI-sourced revenue from everything else, or the GEO number stays out of next year's plan. I'll fund the thing I own. I'll pass on renting visibility to nudge a metric nobody can trace to a sale.

Where they part ways

The Market Analyst and the Skeptic split on timing. The Analyst sees budget already moving to guaranteed placements; the Skeptic says the transaction volume isn't there to justify moving anything yet. Both can be right: retail media wins regardless, because it was winning before AI showed up, and AI is just one more reason on the pile.

The Operator and the CFO agree GEO dies at review, but for different reasons that matter. The Operator kills it because it's unattributable. The CFO kills it because owned media does the same job and he keeps the asset. That distinction decides what survives the cut: the content team gets funded, the GEO vendor does not.

The quiet agreement across all four: brand.com is the real lever, and it's the one nobody is resourcing because it's boring. No slide says "we bought a content team." Every slide says "we're winning at AI."

What this hinges on

One belief: does AI-sourced product discovery drive enough purchases to justify a new spend line this cycle? The evidence in this story says no, not yet. Droesch's own 18-to-20 example is the argument against his own alarm. So the near-term call is cheap: run an incrementality test on owned-media content, skip the standalone GEO package, and let the retail media reps make their pitch because guaranteed placement is a real product with a real invoice you can measure.

The council leans hard one way. Owned media over rented visibility. Retail media over GEO. The compression is real; the vendor answer to it mostly isn't.

Prediction: No large agency holding company (Omnicom, WPP, Publicis, Dentsu, Havas, or Stagwell) will report GEO or "AI visibility optimization" as a distinct, quantified revenue line on any 2026 earnings call, through the Q4 2026 reports in February 2027.

Confidence: Medium. GEO is real work, but too small and too blended to break out.

Why: Droesch's own data undercuts the category: the best lever is brand-owned websites, which is a content-and-SEO job the holdcos already sell, not a new product, and the visibility lift he cites (18% to 20%) is too small to survive a CFO review as a standalone fee. When a service is both hard to attribute and easy to fold into existing content and search practices, agencies bundle it rather than break it out, because a small, un-attributable line invites exactly the scrutiny that kills it. The opposite outcome, a holdco proudly quantifying GEO revenue, would require the category to be big enough to boast about, and nothing in this story says it is. This is a call about the business staying too small to name, not about reporting mechanics: the silence tracks the size.

Revisit by 2027-02-28: We're right if none of the six named holding companies reports a quantified GEO or AI-visibility-optimization revenue figure through their Q4 2026 earnings calls. We're wrong if any one of them puts a specific GEO revenue number in front of investors in that window.