Yelp's AI Bet Is a Real Call Option Investors Are Mispricing

I had Yelp’s Q2 deck open on one monitor and the Fall Release teaser sitting open on the other when an operator I respect — a four-unit casual group on the East Coast — asked me, off the record, what I’d pay for the equity here. The honest answer surprised me, because I have spent most of 2025 telling private buyers not to pay AI premiums for operating groups. The honest answer was: at sub-$3B, the public market is mispricing what Yelp is becoming, and the call option is real.

The thesis, plainly. YELP is not a directory with an AI bolt-on. It is, increasingly, a SaaS-on-content business with 512,000 paying advertisers, an AI feature stack growing by tens of features per release, an operator product line about to add at least two new SKUs at $99–$149/month, and a usage curve on its core Assistant feature that grew more than four-fold year over year. The Street is still drawing the comp set from publishing and local-search. The right comp set sits much closer to vertical SaaS. Until the comp set updates, this is a mispriced call option — not a screaming buy, but the kind of asymmetric setup I want in a thesis-driven book.

I will get four things wrong before this lands. I am going to flag the judgment lines. The math under them is, I think, defensible.

What Q2 actually told us

The single most important sentence in the Q2 2025 SEC filing did not get the attention it deserved. In its 10-Q for the quarter ended June 30, 2025, Yelp wrote that “Yelp Assistant maintained strong momentum, as project submissions through this tool increased by more than 400% year over year.” That is one data point doing a lot of work, so let me say what I think it does and does not tell you.

What it tells you: a feature that did not meaningfully exist as a consumer product two years ago is now generating project submissions — i.e., demand handoffs from a consumer in search of a service to a Yelp-listed business — at more than five times the run-rate of the comparable quarter in 2024. Project submissions are not vanity. They are the action layer that converts a directory pageview into something a paying advertiser actually wants — a qualified lead. Yelp’s request-a-quote machinery has historically been the closest thing the company has to a marketplace conversion event.

What it does not tell you: how much of the +400% is from a small base. A feature that did 10,000 submissions in Q2 2024 and 50,000 in Q2 2025 is up +400%, and that is a real curve. A feature that did 10 and did 50 is also up +400%, and that one is a press release. Yelp has not published the absolute denominator. I have been around tech disclosure long enough to assume the absolute number is meaningful but not yet large — large enough to disclose, not yet large enough to be a P&L line by itself. Judgment, flagged.

The right read, I think, is that Yelp has demonstrated product-market fit on the Assistant surface — consumers are using it, at a growth rate that is unusual even for the AI cohort — and the question that remains is whether Yelp can convert that surface into paid-advertiser revenue. Which brings us to the Fall Release.

What the Fall Release is anticipated to do

Yelp’s Fall product release is expected to drop early next week, on or around October 21. The release will land after this column publishes, so I am not going to recap what they ship — I will write that piece separately, once the press release is in hand. What I want to do here is underline what we already know from the leadup, and price the option accordingly.

The expected shape of the release, based on Yelp’s investor commentary on the Q2 2025 earnings call and the pattern across Yelp’s product release archive, is more than thirty AI features rolled in a single drop, with the headline operator-facing products being Yelp Host and Yelp Receptionist. The market understands Yelp Host as a $149/month AI front-of-house product, with a $99/month price for operators who already pay for Yelp Guest Manager. Yelp Receptionist — the version pitched at service businesses outside restaurants — is expected at $99/month.

If those expected prices land, the unit economics start to look different from how the equity is currently quoted. Three illustrative slices of the same math:

• 5% of Yelp’s restaurant-segment locations adopt Host at a blended $120/month: that’s a couple of thousand seats on the restaurant Guest Manager base, somewhere in the order of $3–4M of incremental ARR. Small in isolation. Meaningful as a category signal.
• 1% of Yelp’s non-restaurant paying advertiser base adopts Receptionist at $99/month: on a base of roughly half a million paying advertisers, that’s about 5,000 seats, ~$6M ARR, with gross margin closer to software than to local-ad sales.
• 0.5% of the full paying-advertiser base adopts either SKU within the first year at a blended $110/month: ~2,600 seats, ~$3.4M ARR.

None of those are large in isolation. The point is the direction of travel: Yelp is converting parts of its base from a CPC advertising relationship into a fixed-fee SaaS relationship. That is not a directory move. That is a hospitality-tech move. The comp set should follow.

I will give you the steelman: 35 AI features in a single release is a lot of features. The historical hit rate on big-bang feature drops in this segment is under 30%. Most of those 35 are going to be incremental — review summaries, response coaching, a couple of new badges. The ones that matter are Host and Receptionist, because they are the first Yelp products that put a fixed-fee SaaS line item in front of an operator. We covered the feature inventory in a forthcoming May piece on Yelp’s AI stack — that companion read is the operator’s-eye view; this column is the investor’s.

Judgment, flagged: I think Host and Receptionist are the two SKUs to watch. I do not think the other ~33 features rerate the equity. They reduce the churn slope inside the existing book, which matters, but does not move the comp set.

The CPC line nobody is reading

Here is the part of the Yelp story that I think is genuinely overlooked. CPCs in the relevant Yelp categories were up roughly 14% year over year in Q3 2025 commentary, against a backdrop where most local-search ad networks were flat to slightly down. That is a price signal. It says the supply of paying advertisers is competing harder for the same auction slot.

Why does that happen? Two things, both of them legible from the Q2 filing. First, the Assistant surface is changing which impressions monetize. A consumer who comes in via “find me a tile contractor for a 200-square-foot bathroom” is a higher-intent click than a consumer who arrives via a generic browse. The Assistant is concentrating intent into auctions where intent was previously diluted. Second, the operator base is growing — Yelp disclosed 512,000 paying advertising locations as of mid-2025 — and the new entrants disproportionately want lead-gen, not branding, which pushes their bid behavior toward the intent-concentrated inventory.

If you take +14% CPC and hold impressions flat, you get a low-double-digit revenue tailwind on the advertising segment that compounds over the SaaS line described above. That is the thesis. Two engines, one of them reflexively connected to the other: the AI surface concentrates intent → CPC rises → operator ROI on Yelp ads improves → operator-side appetite for Yelp’s other products (Host, Receptionist, Guest Manager, etc.) rises, because the platform now has measurable lift inside one product line.

You do not need to believe all of that to underwrite the equity at current prices. You only need to believe that one of the two engines is real.

Where the equity is priced today

Yelp’s market cap sits below $3B as I write this. Net cash is meaningful — the balance sheet has historically run with several hundred million in cash and no debt, and management has been a consistent buyer of the stock through the repurchase program. Strip the cash and you are paying low-2s billions for the operating business.

Against that, you have ~$1.3–1.4B of TTM revenue, a paying-advertiser base of 512,000 locations, an Assistant feature with a quarter-over-quarter +400% YoY usage curve, a Fall Release expected to add at least two SaaS SKUs at $99–$149/month, and a CPC line that is up roughly 14% in a flat local-ad market. The implied EV/Sales is somewhere around 1.6–1.8x trailing, and the implied EV/EBITDA, depending on how you treat one-time items, runs in the 8–11x range.

Now run those multiples against the SaaS-adjacent comp set. Toast trades at roughly 2.4x trailing sales with thinner margins than Yelp prints today. The Olo take-private cleared at roughly a 65% premium to its unaffected close, which I read as a private-market signal about where this category clears when a strategic sponsor underwrites the AI optionality. Lightspeed runs around 3x annualized sales. None of these comps are perfect — Yelp has more advertising exposure, Toast has more transaction exposure, Olo has more enterprise exposure — but they are closer to Yelp’s evolving business model than the publishing and local-search comps the Street still uses.

I am not arguing Yelp deserves a Toast multiple tomorrow. I am arguing the gap between Yelp’s current multiple and the SaaS-adjacent cohort is wider than the differences in business model justify, and the gap is being held open by a comp-set lag that the Fall Release plausibly closes.

Judgment, flagged: I think the right framing is a 12–24 month rerate window contingent on Host and Receptionist landing at meaningful seat counts. That is an option, not a certainty. Price it like one. I would be a buyer of asymmetric upside here. I would not be a buyer of leverage.

What would change my mind

Three lines I am watching. If any of them moves the wrong way, the thesis weakens fast.

Adoption of Host and Receptionist below 1% of the relevant base within four quarters. If, by mid-2026, Yelp has not disclosed seat counts that imply at least the small-end illustrative slice above (~$3–4M ARR), the SaaS-on-content thesis is not landing. I would step out.

CPC reversion. If the +14% CPC line compresses back toward 0% in 2026, the reflexive engine I described is broken. That would tell me Assistant is concentrating intent but not converting it into auction pressure — and that the AI surface is therefore a cost line, not a revenue line.

Repurchase pace. Yelp’s management has historically returned capital aggressively. If buyback pace slows materially without a corresponding M&A move, I read that as a vote of internal confidence I don’t share. The capital-return mechanic is part of the floor on the equity at these multiples.

For the contrarian read against paying any AI premium, see an upcoming May piece where I argue, at length, that operator-side buyers should not pay tooling premiums. The Yelp position is the inverse — you are buying the vendor, not the operator who adopts the vendor — and the math points the other way for that reason alone.

What I’d actually do

Three concrete moves, sized for a thesis-driven book:

• Initiate at sub-$3B market cap, sized as a starter position. Quarter to half of full conviction sizing. The Fall Release next week is the catalyst that either confirms or kills the thesis. There is no reward for being early past it.
• Add on a Host/Receptionist seat-count disclosure in the next two quarterly cycles. A first concrete adoption print is the data point that reprices the comp set. Until that print, the equity is a call option. After it, the equity is either a SaaS-on-content business with real seats, or a directory with a failed product line. The Q4 2025 and Q1 2026 calls are the relevant filings.
• Pair against a long-tail local-search short. Not naming names. The trade is a long Yelp / short generic local-search bet on AI-driven intent concentration. If Yelp’s AI thesis is right, the rest of the long-tail directory cohort gets squeezed. That structure absorbs a big chunk of the macro-ad-cycle risk.

I will give you my honest priors. I think Host clears at meaningful seats. I think Receptionist underperforms expectations because the cross-vertical sell into non-restaurant verticals is harder than the deck makes it look. I think the +400% Assistant number is real but the absolute denominator is smaller than the bull case assumes. And I think the equity still rerates, because the direction of the business model has changed and the comp set has not caught up.

If you only take one thing from this column: Yelp is no longer a directory story. It is a SaaS-on-content story with an AI distribution surface, 512,000 paying advertisers, and a Fall Release next week that anticipated to add at least two fixed-fee SaaS SKUs to the catalog. The Street is priced for the old story. I am positioning for the new one.

The market gets a chance to re-rate on Tuesday. I will be watching for the price-per-seat language in the press release, the seat-count disclosures on the call after, and the CPC line in the Q3 print. If two of three land, the call option is in the money.

— Oliver writes The Bottom Line for TableTransfers. Tips: ma@tabletransfers.com.