The order link on your Google profile isn't yours by default

Open your restaurant's Google listing on a phone. Somewhere near the top there is an order button. Tap it and see where it goes.

For a large share of independent restaurants, it goes to a delivery marketplace taking 15–30% of the ticket — on a customer who already searched for the restaurant by name. That is the most expensive misconfigured field in local search, and almost every guide to Google Business Profile skips it entirely in favor of telling you to post weekly and respond to reviews.

How that link got there without you

Google's documentation is unusually direct about this. Third-party providers "who state they have authorized relationships with your business are automatically listed on your profile."

Read that twice. The provider asserts the relationship. Google does not ask you to confirm it first. It is an opt-out system wearing the clothes of an opt-in one — which is why operators who signed a marketplace agreement years ago, or whose location was simply swept into a provider's feed, find a commission-bearing link occupying the highest-converting spot on their profile.

Three things follow, and each one surprises people:

You cannot delete the link yourself. You submit a request. The provider then has five business days to remove it. If they don't, your recourse is to file a violation report with Google. This is a process with a clock and a counterparty, not a settings toggle.

Setting a "preferred" provider does not remove the others. It adds a badge and gives your link priority placement — and you set it separately for pickup and for delivery. The competing links stay visible. Operators pay ordering vendors for "Google setup" believing they bought exclusivity. They bought a badge.

Turning off ordering to kill the aggregator kills your own link too. There is a profile-level "accept orders" switch, and Google is explicit that flipping it off hides any third-party order options and links you added — yours included. It is the one lever that looks like a solution and is actually a self-inflicted wound.

The correct sequence is: add your own ordering link, set it as preferred for both pickup and delivery, then request removal of the providers you have no current agreement with, then check again in a week. I walked through what those commissions actually cost in the third-party delivery breakdown — this is the same money, leaking from a field most owners have never opened.

Features Google turned off that guides still tell you to use

This is where most published advice is not merely generic but wrong, and it is worth auditing anything you're paying for against this list.

Questions and answers: discontinued 3 November 2025. Google's own developer changelog says it plainly — you can no longer read or post through the API, and the public feature phased out in December, replaced by Gemini-driven answers inside Maps. If an agency is billing you for "Q&A seeding and monitoring," they are billing for work on a feature that does not exist.

Chat and call history: removed 31 July 2024. New chats stopped that July; the export window closed at the end of August.

End-to-end food ordering: shut off 1 July 2024. Customers can no longer build a cart and pay inside Google. Every order button is now a redirect to somebody's website.

That last one comes with a detail I'd normally assume was my mistake. As I write this, Google's own help page still tells restaurants they can accept orders directly through their Business Profile. The feature has been off for two years and the documentation hasn't caught up. I mention it not to score a point but because it is the practical reason so much bad advice persists: people are reading the manufacturer's current documentation and getting a two-year-old answer.

What actually moves the local pack

The best-documented data here is Whitespark's 2026 Local Search Ranking Factors, released 6 November 2025: 47 local-search specialists scoring 187 factors. Treat it for what it is — expert opinion with a stated method, not a measurement of Google's algorithm. Nobody outside Google has the latter. With that caveat, the group weightings for the local pack land at roughly 25% profile signals, 20% review signals, 18% on-page, 15% links, 12% behavioral, 6% citations.

The individual factor scores are more useful than the groupings, and one result should change what you do on Monday:

Being open when someone searches scores higher than your star rating and higher than your review count. For a restaurant that means holiday hours, temporary closures, and the kitchen-closes-at-9-but-the-bar-runs-till-11 problem are ranking inputs, not housekeeping. Accurate hours are cheaper to fix than reviews are to earn, and on this evidence they matter more.

The other lesson from that table is a ceiling. Proximity scores 225 and you cannot optimize it. A restaurant outside the searcher's radius does not enter the three-pack no matter how good the profile is. Any promise to "rank you in the map pack" that doesn't mention geography is selling something.

The name trap: both things are true

Keywords in the business name score 223 — third highest on the list. Google's guidelines also prohibit taglines, phone numbers, URLs, and location descriptors in the name field, and say directly that including unnecessary information "could result in the suspension of your Business Profile."

So: it works, and it can take your listing off Search and Maps entirely. For a restaurant pulling a meaningful share of covers from discovery search, a suspension is a revenue cliff inside 48 hours, and reinstatement requires documentation most owners don't have filed — utility bills in the business name at the address, storefront photography, operating licenses. Most articles quote the ranking value and omit the penalty. You need both numbers to make the decision, and with both numbers in hand it isn't a close call.

AI recommendations are a different game entirely

If you want the genuinely new thing in 2026, it's that visibility in Maps and visibility in AI-generated recommendations have come apart.

Local Falcon published an index in May 2026 covering 10,000 US restaurants across all 50 states, 30 metros, nine grid points per location. It is vendor-produced research — they sell rank tracking — but they publish sample, geography, dates, and query construction, which is more than almost anyone else in this space does. Judge it accordingly.

Their findings are strange in a useful way. McDonald's appeared in 0% of AI restaurant recommendations despite 96% visibility on Google Maps. Independents were recommended at 29.1% versus 13.9% for chains. A restaurant with only social accounts and no website was more AI-visible (39.4%) than one with a conventional website (26.2%). And in searches that explicitly asked for highly-rated places, the strongest band was 4.5–4.7 stars at 40% visibility — near-perfect ratings surfaced less often, apparently because thin review histories read as unproven.

I don't think anyone can tell you a reliable playbook for this yet, and I'd distrust anyone who claims one six months in. What I'd take from it is narrower: the signals that make you findable in Maps are not automatically the signals that make you recommendable by an assistant, and being an independent appears to be an advantage rather than a handicap for the first time in the history of local search.

Where profiles actually break

Relocation. If you move, Google's instruction is to mark the existing profile closed and create a new one — not to edit the address in place. Editing it creates a mismatch against the underlying map entity, and review history is what gets caught in the middle.

Duplicates. One profile per physical location, full stop. A second listing for the catering side or the bar at the same address violates the duplicate rule, and the practical damage is that reviews and photos split across two entities so neither ranks.

Shared kitchens and virtual brands. The address on a profile has to be the place where the business actually operates and holds the space — no booth renters. Several virtual brands at one commissary address is the most common route to a restaurant-category suspension, and there is no clever workaround. If you run virtual brands, budget for this being a permanent constraint rather than a problem to solve.

Your reporting changed too. The performance report no longer gives you the direct-versus-discovery split. It gives you actual search terms, refreshed once at the start of each month, plus views, directions, calls, website clicks, bookings, and menu views. Any dashboard still showing you "% discovery searches" is reconstructing a metric Google retired.

Numbers I deliberately didn't use

Writing this, I discarded more statistics than I kept. "Photos earn 42% more direction requests." "Ten or more photos doubles engagement." "Recent posts drive 24% more interactions." "The average profile generates 59 actions a month." "One additional review generates 600 impressions and 80 visits."

Every one of those circulates constantly and none carries a sample size, a date, or a method. Several trace back to marketing copy that no longer exists on any Google page. They are quoted because they are quotable.

The honest version is duller: complete the profile, keep hours genuinely accurate, get reviews steadily rather than in bursts, respond within a few days — 63% of consumers expect a reply inside a week, per BrightLocal's 2025 survey of 1,026 US adults — and ask for the review fast, because food and beverage customers expect the request same-day or next-day more than any other category measured. None of that needs a fabricated percentage attached to be worth doing. I went deeper on the review side in the reputation post.

The one thing worth doing today

I work at Katalyst, and first-party ordering is something we sell, so weigh that. But the action I'd take doesn't require buying anything from anyone: open your profile on a phone, tap your own order button, and find out who is currently collecting the commission on customers who searched for you by name. If the answer isn't "me," you now know the mechanism, the five-day clock, and the toggle not to touch.

Related Katalyst products

• See Katalyst online ordering