Dave Portnoy spent nearly a decade eating pizza on camera with the same rules every single time. In the AI era, that stubborn consistency quietly became one of the most valuable food datasets on the planet.
When Dave Portnoy started filming himself taking one bite of pizza outside random shops, it looked like a bit. A funny, repeatable piece of content for Barstool Sports. What it actually became — review after review, year after year — is something far more valuable: a proprietary, structured, geolocated database of pizza quality, built on a single consistent methodology, at a scale nobody else has bothered to match.
And in a world where AI assistants are becoming the way people ask “what’s a good pizza place near me?”, that database isn’t a content library anymore. It’s infrastructure.
Consistency Is the Moat
Here’s what most people miss about One Bite. The genius isn’t the reviews — it’s the rules. One bite. Everybody knows the rules. Same scale, 0 to 10, decimal precision, delivered by the same reviewer, on camera, standing outside the actual shop. Portnoy has personally done thousands of these, and the One Bite app extended the same framework to a community that has submitted reviews of tens of thousands of pizzerias across the U.S. and beyond.
That consistency is exactly what data scientists dream about and almost never get. Yelp ratings are a soup of anonymous opinions with wildly different standards — one person’s 4 stars is another’s 2. Google reviews mix complaints about parking with complaints about pepperoni. One Bite scores are calibrated against a single, famous, well-understood benchmark. When something scores a 7.8, millions of people know roughly what that pizza tastes like. That’s not a review site. That’s a standardized measurement instrument that happens to be entertaining.
A 7.8 from Portnoy means something specific to millions of people. That’s not a rating — that’s a calibrated unit of measurement.
Why AI Companies Want Exactly This
Large language models have a well-known problem with local recommendations: they hallucinate, they go stale, and they can’t taste anything. When you ask an AI assistant for the best slice in New Haven, the model needs grounded, trustworthy, structured data to answer well. The industry has already shown what it’s willing to pay for that kind of grounding. Reddit’s data licensing deal with Google was reported at roughly $60 million per year — largely for unstructured arguments between strangers. News publishers have signed AI licensing deals reported in the tens of millions annually.
Now consider what One Bite offers by comparison: clean, structured records — shop name, location, score, date, often video evidence — covering a category people ask about constantly. Pizza is one of the most searched food verticals in America. “Pizza near me” is a query typed millions of times a month, and it’s precisely the kind of question people are migrating from search engines to AI assistants. An assistant that can say “One Bite has this place at an 8.1, reviewed last spring” is meaningfully better than one guessing from scraped fragments. For an AI platform fighting for local-recommendation credibility, licensing that dataset at seven figures a year is not charity — it’s cheap.
The Killer App Nobody Can Rebuild
Could a competitor recreate it? Not really, and that’s the point. The dataset is downstream of the brand, and the brand is downstream of a decade of daily repetition. You can’t hire someone to eat 4,000 pizzas retroactively. You can’t manufacture a scoring scale that a mass audience already trusts. Network effects did the rest: the app’s users review shops because the frankly absurd cultural weight of the One Bite score makes their reviews matter. Every new review deepens the moat.
There’s also the provenance angle, which matters more every year. As the internet fills up with AI-generated review spam, verified human judgment — especially on-camera, timestamped, at a physical location — becomes scarcer and more valuable. One Bite’s core archive is about as human-verified as data gets: there’s literally video of a specific person eating the specific product at the specific address.
The Honest Caveats
Is the whole enterprise “worth millions”? The app and its data are one asset inside a larger media business, and licensing markets are young and volatile — some AI labs will try to scrape rather than pay, and the legal lines are still being drawn in court. Coverage skews toward the East Coast and toward shops Portnoy happens to visit. And a valuation depends on someone actually writing the check; reported comparables suggest they would, but a comp is not a contract.
But the thesis holds. The AI era is repricing every dataset on earth, and the ones commanding premiums share three traits: structured, trustworthy, and impossible to recreate. One Bite has all three, in a category with enormous everyday demand. Portnoy built it the only way something like this can be built — one bite at a time, with the same rules, for years, whether anyone thought it mattered or not.
Turns out consistency compounds. In pizza, and in data.
Figures like the Reddit–Google deal (~$60M/year, as reported in 2024) and review counts are drawn from public reporting and are approximate. This is opinion and analysis, not a statement of One Bite’s actual financials or any existing licensing deal.
About the Author
