Copilot is not really a separate visibility problem. It is Bing with a conversational front end, and almost everything it says about your restaurant can be traced back to a Bing Places listing or a page in Bing's index.
That is good news, because it makes the work concrete and finishable. There is no strategy to develop. There is a listing to claim, a handful of facts to make consistent, and then you are done for a year.
The traceability is the whole advantage
Ask Copilot when a restaurant closes and it tends to show you where it got the answer. Run the same question against ChatGPT or Gemini and you often get a confident sentence with no visible provenance.
For an operator, that difference matters more than which assistant has more users. A cited answer is a debuggable answer. If Copilot tells someone your Tuesday hours are 11 to 9 and you closed Tuesdays two years ago, the citation usually points at the specific page still carrying 11 to 9, and now you have a task instead of a grievance.
Use it that way. Copilot is the cheapest diagnostic tool available for finding stale information about your restaurant on the open web, and it works even if nobody in your market uses Copilot to pick dinner. The wrong pages it surfaces are feeding other systems too, quietly, without telling you where they got it. That is the argument in how restaurants show up in AI search: the sources are shared even when the assistants are not.
Spend an hour asking Copilot the ten questions your staff answers most on the phone. Collect the citations. That list of URLs is a better audit than most paid local SEO reports, and you produced it yourself.
Bing Places is twenty minutes and most people skip it
Microsoft maintains a business listing product analogous to Google's, and the import path from Google Business Profile does most of the data entry for you. Claim, import, verify, check the phone number, check the hours, save.
The reason to bother is not Bing's search share. It is that an unclaimed listing does not sit empty. Microsoft populates it from aggregated data of unknown vintage, and that generated version is what Copilot reads. A restaurant that moved, changed its number, or added Sunday brunch three years ago is often still described by whatever snapshot got scraped before any of that happened.
Verification usually runs by postcard or phone call and can take a week or two, which is the only slow part. The hours and phone fields are what to check most carefully after import, because import maps most fields cleanly and then leaves special hours and holiday closures behind. The detail work is the same as elsewhere, and Bing Places for restaurants walks through the fields that actually change answers.
One thing worth doing while you are in there: make the phone number match the one on your website, your Google listing, and your printed menus, character for character. Formatting differences do not matter much. Different numbers do, and multi-location groups accumulate them the way a kitchen accumulates ladles. NAP consistency for restaurants covers why that one field causes disproportionate trouble.
What Copilot cannot get from a listing
Anything with detail. Whether you have a vegan entree, what the catering minimum is, whether the back room seats twenty, what a large pizza costs on a Tuesday.
For those, Copilot goes to pages, and it will use whatever page it finds. If the highest-quality page describing your menu is a delivery aggregator with a menu from before your last price change, that is the answer people get, and the price will be wrong in a direction that annoys them at pickup.
The fix is unglamorous and it is the same fix as for every other assistant. Put the menu on your own site as text on an HTML page. Put your delivery boundary in a sentence a human could read aloud. Put the catering minimum in writing. Put your holiday closures somewhere other than an Instagram story that vanished in a day.
There is a second-order effect here that operators tend to miss. When your own site carries the clearest version of a fact, the aggregator page stops being the best available source, and it gets cited less. You cannot delete a third-party page, but you can outclass it, and outclassing it is mostly a matter of writing the answer plainly on a page you own. A sentence reading "we deliver within three miles of the Fairview store, with a twenty-five dollar minimum, and delivery ends at 9:30" is more useful to a retrieval system than three paragraphs of description around the same facts.
Directory pages you no longer control are the harder half of this. Some can be claimed and corrected, some can only be reported, and a few will simply persist. Working through them in order of how prominently they rank is the tedious part of citations and directories for restaurants, and it is genuinely tedious. Do the top five and stop.
Where the effort actually pays
Here is the honest accounting. In most American markets, the number of dinner decisions made through Copilot is small. If you are choosing between fixing your Bing listing and fixing the fact that nobody answers your phone between 6 and 8 on a Friday, fix the phone, and it is not close. The volume difference is enormous, and calls are people who already chose you.
The reason to do the Bing work anyway is that it is bounded. One afternoon, no ongoing maintenance beyond an annual check, and the same corrections improve Bing's local pack, Windows taskbar search, and any assistant that leans on Microsoft's index this quarter. It also removes wrong facts from the open web, which raises the odds that any model, asked about you, finds sources that agree with each other. Agreement is most of what makes a model state something confidently rather than hedge.
There is also a category of business where Copilot is worth more than the raw share suggests. Corporate catering ordering happens on work computers, inside Microsoft accounts, often with Copilot sitting in the same window as the email being written. If a meaningful part of your revenue is office lunch drops and boardroom platters, the person placing that order is more likely to be in Microsoft's ecosystem than the average diner picking a Friday dinner. That does not make it a large channel. It does mean the catering page on your site, with the minimum and the lead time written out, is the page most worth having in Bing's index.
Do it in this order: claim Bing Places, fix the phone number everywhere it disagrees, get the menu onto an HTML page, then correct the top few directory listings. Then leave it alone and go look at what happens to your call volume when a Copilot or ChatGPT answer sends someone to your number. The ChatGPT recommendation question has the same shape and the same ceiling.
Set a calendar reminder for a year out to re-run the same ten questions. If the answers still match reality, the work held, and you can skip the rest of the industry's advice on this subject.