How ChatGPT and Google decide which café to recommend

A couple comes out of the metro at Fontana, tired and a little lost, and one of them says to their phone: “where’s a good coffee near me?” Thirty years ago they would have just walked until something looked nice. Today a machine answers first, and it names three places. The question every owner should be asking is simple. Why those three, and not yours?

The honest answer is that nobody at Google or OpenAI has ever been to your café. They have never tasted your cortado or met Jordi behind the bar. They decide entirely from what they can read about you online. So the whole game is making sure what they can read is accurate, specific, and recent.

Here is what they actually read, in plain words.

Your Google Business Profile is the source of truth

Before anything clever happens, there is a boring and very important file: your Google Business Profile. The name, the category, the address, the opening hours, the photos. Both Google and most AI assistants lean on this first, because it is structured and Google trusts it.

If your category says “Restaurant” when you are really a specialty coffee bar, you have quietly told the machine to recommend you for the wrong searches. If your hours are wrong, you get shown as closed and skipped. None of this is glamorous. All of it decides whether you are even in the running.

Reviews are the closest thing to a reputation

Here is where it gets interesting. A machine cannot taste your food, but it can read a hundred people describing it. So reviews are treated as the nearest available proxy for “is this place actually good, and good at what?”

It is not only the star rating. The words matter enormously. When fifteen different guests mention “the carrot cake,” “the terrace in the sun,” and “a quiet spot to work,” you have, without planning it, told every search engine on earth exactly what you are for. Now when someone asks for “a quiet café to work in Gràcia,” the pattern lines up and your name comes out.

This is why a steady trickle of fresh reviews beats a big pile of old ones. Recent reviews tell the machine the place is alive and the picture is current. A café with forty reviews from this year often outranks one with three hundred from 2019.

Your replies are read too, and most owners waste them

This is the part almost nobody knows. When you reply to a review, that reply is public text attached to your business, and it gets read right alongside the review itself.

A reply that says “Thank you!” adds nothing. A reply that says “So glad you enjoyed the flat white and found a calm corner to work, the terrace is quietest before noon” does something real. It confirms the detail the guest raised, in natural language, from the business owner. You have reinforced exactly the thing you want to be known for, and you have done it for free, in a sentence.

Multiply that by every review, in the guest’s own language, and you are feeding the machine a clean, consistent, human description of your café, month after month. That is the single most underused marketing channel a small place has.

So what are the “keywords” for a café?

Forget the word keyword for a second, because it makes people do strange things, like stuffing “best coffee in town” into everything until it reads like a robot wrote it. That actively backfires now. Machines are good at spotting it, and so are people.

The real keywords for a café are just the true, specific things about you:

• The dish or drink people actually come for. The paella, the oat cortado, the pistachio croissant.
• The neighborhood, named plainly. Gràcia, El Born, Blloku.
• The occasion you are good for. Working in the morning, a first date, brunch with the kids, a quiet glass of wine.

If those three things show up honestly across your profile, your reviews, and your replies, you will get recommended for the searches that fit you, to the people who were going to love you anyway. That is the whole trick. It is less about being clever and more about being clear.

A menu the machine can actually read

One last piece people miss. A menu that lives only as a PDF or a photo is almost invisible to a search engine. A menu published as real text, with dish names and prices it can parse, is one more honest, structured signal about what you serve and what it costs. It helps you show up for “vegan brunch near me” in a way a flat image never will.

The uncomfortable summary

You are already being described to strangers by a machine that has never visited you. Right now that description is built from whatever happens to be online, which for most cafés means a thin profile, a few old reviews, and either silence or a row of “Thanks!” under them.

You cannot make the machine like you. You can make sure it knows, accurately and recently, what you are genuinely good at. Keep the profile honest, keep real reviews coming in, answer them like a human who was actually there, and publish a menu it can read. Do that, and when the couple at Fontana asks their phone, there is a real chance it says your name.