Yelp Ships AI Voice Agents for Restaurants and 15 Guest Manager Updates

I was halfway through a lukewarm cortado at a coffee bar on Valencia Street this morning when the Yelp Spring Product Release dropped into my inbox, and the PR copy did what PR copy always does: it buried the lede. Fifteen new features, a cheerful CPO quote, a screenshot of Yelp Assistant identifying a plate of cacio e pepe from a photo. I scrolled. Photo recognition. AI response quality badges. Traffic Attribution Dashboard. Guest Experience Surveys for Yelp Guest Manager Plus and Enterprise. Cute. Cute. Cute.

And then, two-thirds of the way down the post on blog.yelp.com, one line: AI voice agents, in test, for restaurants and services. I put the cortado down.

My read: that single sentence is the entire release. Everything else is product table-stakes — the kind of feature backlog any mature marketplace ships every spring to keep operators from churning. The voice agents line is Yelp telling Wall Street, in a tone calibrated to sound humble, that the search-and-discover pure-play is becoming a vertical SaaS company. The targets are OpenTable and SevenRooms, and the timing is not accidental.

The 15-feature drop, ranked

Let me actually rank them, because the press release does not.

Tier one — moves the P&L. AI voice agents in test for restaurants and services. Guest Experience Surveys inside Yelp Guest Manager Plus/Enterprise (because that is the seat-based SaaS line, and surveys are the closest thing to a CRM hook Yelp has shipped). Traffic Attribution Dashboard for advertisers, which is the kind of measurement product that defends the ads business when LLM-based search starts eating top-of-funnel.

Tier two — clever consumer surface, unclear monetization. Yelp Assistant photo recognition on iOS. AI response quality badges. Both are the sort of thing TechCrunch’s writeup and Search Engine Land’s coverage lead with, and both are essentially LLM wrappers grafted onto Yelp’s existing review corpus. Useful. Not a business line.

Tier three — the other ten. Honestly, they are the kind of CRM, messaging, and dashboard polish you ship to keep the multi-product story alive on an earnings call. I read each one. I will not waste your scroll.

CPO Craig Saldanha’s framing, verbatim from the release: “we’re continuing to harness AI to unlock the potential of Yelp’s rich data in ways that build trust and simplify decision-making.” Translation: we have 20-plus years of review data, we are going to package it into AI features and sell those features into the seat-based Guest Manager line. The data moat finally has a price tag attached.

Why the AI voice agents test is the headline

Restaurants do not need another review widget. They need someone to answer the phone. Anyone who has worked a host stand at 6:45 on a Friday knows the call volume is the single most distracting interrupt in the room — reservation changes, “are you open,” “do you have a high chair,” allergen questions, the lost-AirPods caller. A working voice agent that handles even sixty percent of inbound calls is, no exaggeration, a labor-line item. It is also a moat against OpenTable, because OpenTable does not own the phone number; Yelp does, by virtue of being the default business profile a caller lands on after a Google search.

The Yahoo Finance summary noted that voice agents are explicitly framed for “restaurants and services” — meaning Yelp’s plumbers-and-electricians arm gets the same engine. That dual deployment is what makes this a platform bet rather than a feature. The home-services side will train the model on call patterns that look nothing like a restaurant’s, and the cross-pollination is the kind of compounding advantage that, in a later piece on POS-side AI — see what Toast is doing with Sous Chef — I have argued is the actual durable edge in this category.

My read: the test is small and deliberately under-promised. Yelp wants to ship a screenshot at the next investor day, not a metric. If the metric were good, they would have led with it.

What this means for OpenTable and SevenRooms

OpenTable has been quietly defensible because Booking Holdings carries it on a balance sheet that does not require it to grow. SevenRooms is the actual operator favorite — better data model, better hospitality CRM, better at the things Yelp Guest Manager has historically been weak on. Both now have to price in a Yelp that is not just a top-of-funnel ads tool but a full seat-based stack: reservations, waitlist, surveys, voice, and the consumer-side discovery layer feeding all of it.

There is also the rumored DoorDash–SevenRooms tie-up (rumored — nothing announced, and I will believe it when a filing or a press release shows up in early May). If DoorDash does buy SevenRooms, Yelp’s voice-agents test becomes the only independent path for a restaurant operator who does not want to hand the front-of-house stack to either DoorDash or Booking Holdings. That is a real wedge, and it is the kind of strategic positioning that, as our later Vibe Check argues about the broader rebrand-and-reposition cycle in this category, tends to show up first in product roadmaps and only later in S-1 language.

My read: SevenRooms’s product team saw this release this morning and the slack channels lit up. They should be flattered and nervous in roughly equal measure.

What to watch on Yelp’s Q1 print

Yelp reports Q1 in the forward weeks. Three things to watch.

One: any disclosure on Guest Manager Plus/Enterprise seat count or ARR. If they break it out, it means they are about to treat it as a reportable segment, which is the prerequisite for the SaaS multiple they want.

Two: ad revenue trajectory against the Traffic Attribution Dashboard launch. If attribution is improving advertiser retention, the dashboard is doing its job — and Yelp will say so on the call.

Three: any color on the voice agents pilot. I expect a single sentence, a single restaurant name, and zero metrics. That is the right disclosure for an in-test product. If they give a number, it means the number is great and they are pre-marketing the next release.

Wendy’s Q1 lands tomorrow on the QSR side, Hilton printed this morning on the hospitality side — different cycle, separate piece — and the broader read across all three is that 2025 is the year restaurant-adjacent tech stops being a feature line and starts being a P&L line.

The cortado, by the way, was terrible.

— Maya covers restaurant tech. Tips: [email protected].