The marketing lead at a small Bristol chain opened Bing Webmaster Tools on a Tuesday morning and found the trust score on the group's flagship venue page had slipped. The JSON-LD at the bottom of the page still listed the lamb shank as available. The HTML higher up, updated by the kitchen on Friday, marked it sold out. Until last week, that was an oversight. Under Bing's May 2026 update, it is a ranking cost.
Bing rewrote how it weights schema
SEOvendor published a write-up of Bing's May 2026 algorithm refresh on 11 May. Bing now leans harder on structured data — LocalBusiness, FAQ, Service — and weights the alignment between schema markup and the rendered page. The example in the write-up is direct: "If your markup says 'InStock' but an AJAX button shows 'Sold Out,' search engines will lose trust in your data fast."
The instruction is not "add more markup". It is "stop letting your markup drift from the page".
The drift is normal in restaurants
A pub kitchen in Sheffield takes lamb off the menu on Monday. The chef rewrites the printed card on Wednesday. The website's HTML is updated on Friday. The JSON-LD has not been touched in eight months. By Friday evening, the page tells Bing one thing and a customer another.
The same drift sits inside dietary tags, prices, opening hours, and the dishes a venue claims to serve at lunch. The lamb dish in the JSON-LD says suitableForDiet: GlutenFreeDiet. The kitchen swapped the gravy supplier in March and the new stock contains wheat. The site has not caught up. Bing reads the tag as a claim, the page as evidence, and downgrades the data when the two disagree.
Bing notices.
ChatGPT Search and Copilot read Bing
Bing's UK market share is small. The reason a restaurant should still care about its structured-data update is that ChatGPT Search and Copilot both index through Bing. A change to how Bing weights schema is a change to what a substantial share of AI restaurant discovery sees and trusts.
A diner who asks Copilot for a roast near Piccadilly Gardens gets a shortlist assembled from Bing's index. A venue whose schema and page have drifted sits further down that list — not because a human visitor was misled, but because the model's confidence in the underlying data dropped.
Bing's trust score is the data layer ChatGPT Search and Copilot read restaurants through.
Multi-location chains take the sharpest hit
The SEOvendor write-up flags that multi-location businesses face particular pressure under the new weighting. Generic copy across cities sinks rankings fast.
For a twelve-site chain running the same menu page behind a different address field, the trust signal collapses everywhere at once. The fix is mechanical. Each location's schema needs to reflect what that kitchen actually does: hours, the dishes that site runs this week, prices it charges, dietary information specific to its preparation. Where the same dish travels across the estate, the description can repeat. Where it does not, the schema should not pretend.
The cleanup is dull and immediate
The audit is the boring one. Open the menu page. Read what the JSON-LD claims about hours, dishes, prices, dietary tags. Open the page on a phone and check the same facts in the visible HTML. Where the two disagree, fix the source the menu software pulls from.
Menu software that generates Schema.org markup from a single source of truth — GMMO is one option — fails the drift test less often because there is nothing to drift from. The dish is described once. The HTML and the JSON-LD render off the same record.
The lamb shank is back on the menu in July. By then, the JSON-LD on the chain's flagship page will say it is, or it will not, and Bing will read both.