The Jealous Fork belongs to Orbator’s founder and nobody has done AI visibility work on it. That is exactly why it is worth publishing: it shows what these answers do when a business is left alone, measured engine by engine and re-pulled from the Index every time this page rebuilds.
Orbator founder · published Jul 2, 2026 · live data checked Aug 17, 2026
I own this restaurant: the Jealous Fork belongs to Orbator’s founder, and no AI visibility work has ever been done on it. No profile rebuilds, no review campaigns, no placements, nothing. That is the point of this page. It is a control, not a testimonial: it shows what AI answers do to an ordinary business that nobody is working on. Every verdict below is pulled live from the same public AI Recommendation Index that measures every business, and it updates automatically whether the numbers flatter us or not.
Most case studies show you a business that was optimized. This one shows the opposite, and that is the reason it exists. The Jealous Fork is my restaurant. No profiles were rebuilt for it, no review campaign was run, no listings were pitched, no content was written to get it into anything. It has simply been measured.
That makes it a control. Every claim about AI visibility work runs into the same question: would the answer have moved anyway? Here is one business where nobody touched the inputs, sampled across ChatGPT, Claude, Gemini, Perplexity and Grok on the same two questions, over months.
The short version is that the answers moved anyway, in both directions, and the reason is visible in what the engines read.
“Best brunch in Miami” has never named the Jealous Fork. Not once, in any sampled answer, on any engine, since this category entered the Index in June 2026. That is a genuinely flat line across every run below.
The engines are not confused here, they are consistent. This question in this city belongs to a set of restaurants that the local food press has written about for years, and the citation list under this block shows exactly which publications those are.
The flat line. Not named in a single sampled answer since we started measuring this category. A control case has to show the categories that never move.
The same restaurant, a narrower question, and a completely different picture. On 8 and 9 July 2026, four of the five engines named the Jealous Fork for “best pancakes in Miami”. Anthropic did not. Two weeks later, across the three runs sampled on 22 and 25 July, none of them did. On 14 August, with all five engines run in the same session, two named it: Gemini and Grok.
Aggregated by month that reads as 5 of 10 sampled runs in July and 2 of 6 in August. Nothing at the restaurant changed between those dates. No menu change, no campaign, no new listings, no work of any kind.
This is the finding, and it is more useful than a win would have been. A business’s presence in AI answers is not a property you earn once and keep. It is a rate that moves week to week, and it moves differently on each engine.
The category that moves on its own. This block re-checks the Index on every refresh, so whatever it says is the current answer, not our best week.
Because every run stores the pages the engine consulted, the swing is traceable. On 9 July, OpenAI answered this question grounded almost entirely in google.com results, and it named the Jealous Fork. On 14 August, the same prompt on the same engine pulled from a different set, mostly curated local food media, and it did not. The restaurant did not leave those answers. The retrieval set changed underneath them.
The two engines that named it on 14 August are the two reading widest. Grok consulted 46 sources that run, including the restaurant’s own site and the social platforms where its customers post. The three that did not name it were drawing on a narrower diet of local listicles and aggregators, and the Jealous Fork is not in those lists.
One result deserves its own sentence, because it complicates the obvious conclusion. Perplexity cited jealousfork.com in four separate runs, in July and again in August, and only named the restaurant in the July ones. Being read is not the same as being recommended. Anyone selling citations as if they were recommendations is skipping that step.
Gemini is the honest gap in this analysis. Its citations resolve to Google redirector URLs rather than publishers, so for that engine we can see the verdict but not the sources behind it.
The sample sizes are small and we are not going to dress them up. Five of ten runs in July carries a 95% confidence interval of roughly 24% to 76%. Two of six in August lands at roughly 10% to 70%. Those intervals overlap heavily, which means the drop you can see in the timeline is not statistically distinguishable from noise. What is solid is the direction of the mechanism, not the size of the change.
This is observational. Nobody ran an experiment here, nothing was held constant on purpose, and the engines are free to change their retrieval and their models without telling anyone. What we can show is which sources each engine consulted and which answer it gave. What we cannot show is causation.
And to be explicit, because the earlier version of this page got it wrong: no work was done on this restaurant, so nothing here is evidence that any particular tactic works. It is evidence that the answers move, that they move differently per engine, and that a single check on a single day tells you almost nothing.
What we did, dated — interleaved with the Index’s live month-by-month record of how often AI named Jealous Fork for “pancakes in Miami”.
The “best brunch in Miami” category starts being sampled. The Jealous Fork is named in none of the five runs that month, and in none since.
Across the three runs sampled on 22 and 25 July, no engine names it. Nothing changed at the restaurant in between.
On 8 and 9 July, Grok, Gemini, OpenAI and Perplexity all name the Jealous Fork for “best pancakes in Miami”. Anthropic does not.
On 14 August, Gemini and Grok name it, OpenAI, Anthropic and Perplexity do not. Perplexity cites jealousfork.com in the same run where it declines to recommend it.
No. Orbator’s founder owns the Jealous Fork, which is disclosed at the top of the page. The verdicts shown are pulled live from the public AI Recommendation Index, which measures every business with the same methodology — rankings cannot be bought, including by us.
No. None. No profile rebuilds, no review campaigns, no placements, no content written to get it listed anywhere. An earlier version of this page described an optimization programme that was never carried out, and it has been removed. The restaurant has only ever been measured, which is what makes it useful as a control.
It is why the data is live rather than curated. The engine-by-engine verdicts, the “who AI recommends today” lists and the timeline are re-pulled from the Index automatically, on the same methodology used for every other business. The page currently shows a category we have never appeared in and a category where three of five engines decline to name us, which is not what a biased page would look like.
The live blocks refresh automatically. The page re-pulls the current verdict from the AI Recommendation Index roughly every hour, and each block shows when its data was last checked. The narrative is hand-written, the numbers are not.
Mainly that one check is not a measurement. The same question on the same engine gave opposite answers five weeks apart here, with nothing changed at the business, and the five engines disagree with each other on any given day. Before deciding whether you have an AI visibility problem, find out what the rate actually is and how it moves. Start with the free scan: it shows whether AI names you today, who it names instead, and which sources each engine read to decide.
Free, 20 seconds, across the AI engines your customers actually ask. You will see who AI recommends today and which sources each engine read to get there.
RUN MY FREE SCAN →