Yelp's local-ad engine is cracking. The AI pivot won't save Q2.

I read Yelp’s Q1 shareholder letter on Friday morning at the desk, print version, because the screen version always flatters and the print version always tells. It hit the wire after the close on Thursday, May 7 — the Q1 letter to shareholders on the SEC site, the 8-K filing wrap, and the transcript overnight. The sell-side notes were three deep by morning, most running the “AI products are scaling” frame. I read the print version twice and circled five lines in pencil. The AI framing is not the story.

Here is the contrarian thesis before the next call rolls over the narrative: Yelp is a melting-ice-cube on its core Restaurants/Retail/Other ad surface, the buyback is the only thing currently anchoring the multiple, and unless Q2 produces a hard inflection in RR&O the strategic case for a take-private by Q4 opens wide. That is judgment, not fact — flagged here and again later. The math is the math.

The Q1 numbers, plain

Net revenue printed $361.5M, up 1% year-over-year. That is the headline the wire ran. Inside it the picture is uglier. Services advertising revenue — the home-services, legal, auto, professional categories that Yelp has been migrating toward for three years — was $234M, up roughly 1%. Restaurants, Retail & Other advertising revenue — the original Yelp surface, the local-merchant ad engine that built the company — was $99M, down 11% year-over-year. “Other revenue,” the line that captures Hatch (acquired late 2024), data licensing, and the food-ordering rev share, was $29M, up 75% year-over-year, but it is still under 10% of the total.

Adjusted EBITDA came in at $79.4M, down 7% year-over-year, a 22% margin. Net income was $18M, GAAP diluted EPS $0.30, down 27% year-over-year. Paying advertising locations were 485k, down 6% year-over-year. Ad clicks were down 10%. The company repurchased 5.1M shares for $125M in the quarter. They reaffirmed full-year FY26 revenue guidance of $1.455B-$1.475B and Adjusted EBITDA of $310M-$330M — the same range they walked into the year with in February.

Jeremy Stoppelman framed it on the call as a year of “AI-driven product reinvention.” David Schwarzbach, CFO, framed it as “disciplined cost management in a transition year.” Jed Nachman, who runs local sales, talked up Hatch and the food-ordering layer. The framing is internally consistent. It is also what every management team says when the core surface is bleeding and the new surface is too small to plug the hole.

What -11% on RR&O actually means

Restaurants, Retail & Other ad revenue dropped from roughly $111M in Q1 last year to $99M this quarter. That is twelve million dollars of annualized run-rate revenue gone, on a line that was already losing share to Google’s local pack, Meta’s neighborhood ads, and increasingly the in-app ordering surfaces of the third-party marketplaces. Twelve million off RR&O at, call it, an 80% incremental contribution margin — Yelp doesn’t disclose the line-level economics, but local self-serve ad inventory historically clears at that gross level — implies roughly $9M-$10M of EBITDA leakage on the quarter from RR&O alone.

Now layer the paying-locations number on top. Yelp had 515k paying advertising locations a year ago. They have 485k today. That is 30,000 paying merchants gone in twelve months. The company will tell you the mix is shifting — Services merchants pay more per location than RR&O merchants, and the dollar-weighted picture is healthier than the headcount-weighted picture. That is true and it is also irrelevant to the question of franchise durability. The 485k number is the population funnel. If the population funnel keeps shrinking, the dollar-weighted picture catches up to it on a one-to-two-year lag.

Ad clicks down 10% is the leading indicator. Demand for the surface is down 10%, regardless of pricing. You can hold revenue flat on a declining click base for one or two quarters by raising CPC — which is what the services-side numbers suggest is happening — but not indefinitely. Either click volume stabilizes or the price ceiling becomes binding and services advertisers start churning too. That is the melting-ice-cube concern. The demand signal on the surface is decaying, and the question is whether AI-driven product reinvention can rebuild click volume before the price compensation runs out.

I do not think it can. Not by Q4.

The “other revenue” hopium

Let me deal with the bull case on its own terms before I close the take-private argument.

The bull case is that “Other revenue” — the $29M line — grew 75% year-over-year and is the leading edge of where Yelp is going. Hatch is the AI-powered communications product Yelp picked up in 2024; it sits at the merchant-CRM layer and competes with Podium and Birdeye. The food-ordering layer is the Grubhub-rev-share business plus the in-app ordering integrations. Data licensing is the third-party API access for AI training and aggregator partners. Stoppelman and Nachman both spent significant transcript time on Hatch ARR growth and the food-ordering attach rate.

Run it forward. If Other revenue grows 75% Y/Y for four straight quarters — heroic, because the comp gets harder every quarter as Hatch laps its acquisition base — you go from $29M in Q1 to roughly $51M in Q1 next year. Annualized that is $200M-ish, against an FY26 revenue range of $1.46B-$1.48B. Even at that pace Other is 13%-14% of total by mid-2027. Meanwhile if RR&O continues at -11% Y/Y you lose roughly $44M of annual RR&O revenue in the same window. The new line, growing at heroic rates, is barely outpacing the bleed in the old line. That is not a transition. That is a treadmill.

The harder critique is that 75% is flattered by the Hatch acquisition lapping in. Strip out inorganic contribution and underlying organic growth is more like high-twenties — still the right direction, but not the runaway figure the narrative wants you to anchor on. By Q3 the inorganic lap is fully through and you find out what the real rate is. My guess is 25%-35% organic. Good business. Not a franchise-saver at this scale.

The buyback math: $125M Q1 as the floor

The single line in the print that most clearly tells me management knows what I know is the buyback line. Yelp repurchased 5.1M shares for $125M in Q1. At a $3B-ish market cap that is roughly 4% of the float in a single quarter. Annualized, if they keep the pace, that is roughly 16% of the float taken out per year. They will not keep the pace — they cannot, at that rate the share-count math runs into the buyback authorization ceiling by 2027 — but the Q1 pace tells you what the floor under the multiple is.

The mechanic is straightforward: if RR&O bleeds at -10% Y/Y and Services grows at +2% Y/Y, total revenue is roughly flat for FY26 and the reaffirmed guide is achievable, but Adjusted EBITDA compresses 5%-10% because the mix shift is unfavorable to margin. EPS is then held up by share count coming down. The buyback is not creating value. It is masking a deteriorating operating model. That is legitimate when management has conviction in a transition — Apple did this from 2013 to 2017. It is also what management does when they are running out of operating levers and need to keep the multiple intact through a hard period.

The Q1 buyback pace tells me management will deploy capital aggressively to defend the equity. Which is the prerequisite for the next argument.

The take-private case (interpretation flag)

This next section is interpretation and judgment, not fact. Flagging it because the claim gets re-circulated as if sourced. It is not. It is the read.

If RR&O drops below 8% of paying-location revenue mix in Q2 — currently roughly 30% by revenue, but click and engagement metrics show the underlying franchise decaying faster than the revenue line — the strategic logic for a take-private opens in three directions.

One: a financial sponsor takes Yelp private, eliminates public-company compliance costs (call it $25M-$30M annually), runs the buyback math to the natural conclusion (LBO arithmetic at roughly $3.5B equity plus $500M of new debt against $310M-$330M of Adjusted EBITDA gets you to 12x leverage-adjusted, which is sporty but financeable in the current credit window). The Services side becomes the core asset; RR&O gets monetized for what cash it still produces. Hatch and Other become the growth story for an eventual re-IPO or strategic sale in 2028-2029.

Two: a strategic acquirer — and the natural candidates are narrower than people assume. It is not Google (antitrust). It is not Meta (antitrust and category mismatch). It is more plausibly a category-adjacent platform that wants the Services advertiser book and the Yelp brand on the local-merchant CRM side: Angi, ServiceTitan, or even one of the larger restaurant-tech consolidators that has been on the hunt — see my restaurant-tech M&A roundup for the cast of characters there. The Services-side ad book is the prize. RR&O is a divestiture asset.

Three: a hybrid — sponsor leads, a strategic takes a minority position, the carve-out of Hatch and Other revenue gets pre-negotiated. This is the most likely structure if it happens, because it solves the financing problem (lower equity check for the sponsor) and the antitrust problem (the strategic doesn’t take the whole consumer surface).

The timing argument for “by Q4” rests on two pieces of math. First, the sell-side consensus for FY26 is currently at the bottom end of management’s guide. If Q2 misses — and the click trajectory plus the RR&O trajectory says it more likely than not misses — the consensus reset takes the stock down 15%-25%, which brings the equity check on a take-private into a band where sponsors actually engage. Second, the credit window for restaurant-and-local-services LBO paper is open right now in a way it was not for most of 2024-2025; my NRP Florida note walked through that paper market and the Olo take-private piece on Thoma Bravo is the most recent comp for how this category actually clears.

For context on the AI premium that the bull case relies on, see the broader argument in the case against AI premium — tense-neutral on that one because it predates this Q1 print. And the Yelp AI stack post lays out what the company is actually shipping on the product side, which is more than the bears give credit for but less than the bulls are pricing in. The Chipotle Q1 note from two weeks ago is the parallel case study for how operations-and-loyalty depth beats AI-narrative depth in this tape.

What I am watching in Q2

The bet I want to put down for the next 12 weeks: Q2 is the inflection. If RR&O ad revenue runs better than -8% Y/Y, the bull case has a leg. The product reinvention is buying time and the bleed is decelerating and you can argue for patience. If RR&O runs at -11% to -15% Y/Y — which is what the Q1 click trajectory implies if you straight-line it — the strategic case for a take-private opens. If RR&O runs worse than -15%, the only question is whether the board engages a sell-side advisor in Q3 or Q4.

The paying-locations number is the second tell. 485k in Q1. If Q2 prints below 470k — which is consistent with the current attrition rate — the funnel argument becomes incontestable and management loses the “stabilizing” framing they leaned on in this quarter’s call.

The third tell is the Other revenue growth rate ex-Hatch. Management does not break this out. The sell-side will try to back it out and the result will land somewhere between 20% and 35% organic. Anything below 25% organic and the runaway-growth narrative breaks. Anything above 35% organic and the bull case picks up real fuel.

The fourth tell, and the one I am going to be watching the hardest, is the language on the Q2 call about strategic alternatives. Management has not used the phrase. They will not use it voluntarily. But the way they answer the inevitable analyst question — and someone will ask it, because the buyback pace and the RR&O bleed make it obvious — tells you everything about whether the board has already engaged.

Bottom line

Net revenue $361.5M up 1%. RR&O down 11%. Paying locations down 6%. Clicks down 10%. Other revenue up 75% but under 10% of total. EBITDA down 7%. $125M buyback. Guide reaffirmed. AI products invoked nineteen times in the transcript.

The print is not a disaster. It is also not a turn. It is a melting-ice-cube quarter with the buyback acting as the floor under the multiple and the Other-revenue line acting as the bull-case ladder out. The ladder is real but it is too short, and the ice is melting faster than the ladder is being built. By Q4 either management proves the click trajectory is stabilizing — in which case the stock is interesting at the current multiple — or the strategic case becomes the only case, and a take-private becomes the path that solves the math.

I would not be short the stock here, because the buyback floor is real and the strategic optionality is real. I would also not be long here, because the operating thesis requires you to believe AI-product reinvention can rebuild click volume on the core surface in two quarters, and I cannot get there from the Q1 print.

Watch Q2.

— Oliver writes The Bottom Line for TableTransfers. Tips: [email protected].