How to show up in ChatGPT recommendations (and Gemini)

Traditional SEO is about earning a spot on a page of blue links that a person then scans. Generative engine optimization, or GEO, is about being the answer an AI assistant gives when someone asks it a question directly. When a customer in Manila types 'best halal ramen near BGC' into ChatGPT, they do not get ten links. They get a short list of named places, and either you are on it or you are invisible.

The mechanics differ. A search engine ranks pages. A generative assistant reads across many sources, then synthesizes a single recommendation in its own words. You are no longer optimizing to be clicked. You are optimizing to be cited, quoted, and named.

The good news is that GEO and SEO are not enemies. Almost everything that makes you legible to Google also makes you legible to the models. The difference is emphasis: clarity, consistency, and structure matter even more when a machine is summarizing you instead of a person deciding for themselves.

No assistant knows your shop from memory. When you ask about local businesses, these tools reach for live data through search and retrieval, then blend it with what they learned during training. OpenAI says ChatGPT search can send rewritten queries to partner search providers and can use your approximate location for local results. That means the sources feeding them are, roughly, the same public web the rest of us can see.

For local queries specifically, a handful of sources carry most of the weight.

• Your own website, including any structured data (schema markup) on it.
• Your Google Business Profile, still the single richest structured record of a local business.
• Other map and place listings, such as Apple Maps, Bing Places, and Foursquare.
• Reviews and their actual text, on Google, Facebook, and niche directories.
• Third-party directories, listicles, and 'best of' articles that mention you by name.

Perplexity leans heavily on live web retrieval and cites its sources openly, so you can often see exactly which pages it read. Gemini can use public information from Google Maps to find places and give their hours and ratings. ChatGPT blends web search with training data. The precise recipe each one uses is proprietary and shifting, but the inputs overlap enough that a single, well-built local presence serves all three.

That is the strategic relief in all of this. You do not need three separate playbooks. You need one clean, consistent, and complete public footprint that any retrieval system can read without getting confused.

Models are confidence machines. When your name, address, and phone number match everywhere they look, an assistant can state your details plainly. When sources disagree, the model may hedge, pick the wrong branch, or name a competitor it has clearer data on. The Google Business Profile optimization guide is where that single, consistent record starts.

Completeness compounds. A profile that spells out your category, hours, services, prices, and payment methods gives an assistant more true things to say about you, which makes it likelier to surface you for a specific question like 'which clinic in Makati opens on Sunday and takes cards'.

• Pick one exact NAP format and use it letter for letter on every platform.
• Fill every field: category, services, hours, holiday hours, attributes, and a plain description.
• Keep hours and holiday closures current, since stale hours get repeated back to customers.
• Add the details customers actually ask about, such as halal, GCash, wheelchair access, or parking.

If structured data tells a model the facts, reviews and citations tell it the reputation. Assistants can draw on the language in your reviews as well as your star average, and that text is often what they paraphrase back. Reviews that mention specific dishes, services, or neighborhoods give the model concrete reasons to recommend you. How to ask customers for Google reviews is how you get more of that language. Answering them helps too; the free AI review reply generator drafts a reply in the customer's language.

Citations, meaning other credible pages that mention your business by name, matter too. When a respected local blog lists you among the best coffee spots in Cebu, that sentence can become the very sentence an assistant repeats. Being named in third-party 'best of' content gives an assistant something concrete to quote.

• Encourage genuine, detailed reviews rather than one-line ratings.
• Reply to reviews in natural language, which adds relevant, quotable text to your profile.
• Get listed in reputable local directories and rounded-up articles for your city and category.
• Never buy reviews or mentions, since fakes get filtered and erode the trust you are trying to build.

A surprising number of local businesses in the region run entirely on a Facebook Page with no website at all. That works for humans, but it leaves the models thin on structured data they can crawl and cite. A simple, fast, crawlable site is still your strongest owned asset for GEO.

The point is not a fancy design. It is machine-readable clarity. Put your NAP in plain text, add LocalBusiness schema, and answer the real questions customers ask, in the words they use. Schema is good practice, not a shortcut: Google says its AI features need no special markup or AI text files, only a page that is indexed and eligible to show in Search with a snippet.

• Show your name, address, and phone in real text, not baked into an image.
• Add LocalBusiness or Restaurant schema markup with hours, geo coordinates, and price range.
• Write an honest FAQ answering the questions people actually ask an assistant.
• Make sure the site is fast, mobile-first, and not blocking crawlers in robots.txt, including Googlebot and OAI-SearchBot, which ChatGPT search uses.

Any plan for this has to start from how Filipinos actually search. The Philippines is mobile-first, and for many customers the internet effectively begins inside Facebook, Messenger and Viber rather than a browser.

That shapes two things. First, your Facebook Page is often your most-seen listing, so its details must match your Google Business Profile exactly. Second, as assistants get embedded into the apps people already live in, the businesses with clean, consistent public data will be the ones those assistants can confidently name.

• Treat your Facebook Page as a primary listing, not an afterthought, and keep its NAP identical to Google.
• Do not neglect Google Business Profile even in Facebook-heavy markets, since Gemini uses Google Maps data and Google tells businesses to keep their Business Profile current for its AI features.
• Use local languages and the mixed English many customers type, such as Taglish or Bahasa-English.

You do not need to wait for the dust to settle. The concrete steps that help today are the same fundamentals that will still matter when the tools mature, which makes this rare low-risk ground. Most of the list overlaps with how to rank higher on Google Maps, which is the point: one job, two payoffs.

That last step below is underrated. Query the assistants the way your customers would, note who they name and which sources they cite, and you have a free, direct read on your AI visibility. Repeat it every month or two and keep notes, so you can see whether anything changes.

• Audit your NAP across Google, Facebook, Apple Maps, and every directory, then fix mismatches.
• Complete your Google Business Profile to 100 percent, including attributes and photos.
• Publish or refresh a crawlable website with plain-text NAP and LocalBusiness schema.
• Build a steady flow of detailed, genuine reviews and reply to them.
• Earn mentions in trusted local directories and 'best of' articles in your niche.
• Test it: ask ChatGPT, Gemini, and Perplexity for recommendations in your category and city, and see whether you appear.

Honesty matters here, because the field is young and full of confident guesses. What is well established is that complete, accurate details and a strong review profile help you on Google and in maps apps. For AI assistants the evidence is thinner, but the same foundation is the safest bet.

What is still emerging is the precise weighting each assistant gives to each source, how often they refresh, and how much a single directory mention really moves an answer. These change quietly and without announcement. Be wary of anyone selling guaranteed 'rank number one in ChatGPT' packages, because no such guarantee exists.

The sane strategy is to invest in the durable fundamentals now and treat GEO as an extension of good local presence, not a separate dark art. The businesses that show up in AI answers a year from now will mostly be the ones that were already easy for a machine to understand.