Contentoren

SEO & GEO

Local AI Search 2026: Getting Recommended by ChatGPT & Gemini

Local AI search optimization 2026: where ChatGPT, Gemini, Copilot, Perplexity and Apple get business data, plus a NAP audit and a 20-prompt test.

19 min readby Contentoren Editorial Team
Share
A cycling couple in the Spreewald looks on a smartphone for a place to eat, a village inn behind them.

Key takeaways

In short:

  • Each AI assistant takes local business data from different sources: Google (AI Overviews, AI Mode, Gemini) from the Business Profile and Google Maps, ChatGPT from third-party providers that OpenAI only partly names, Copilot from the Bing ecosystem, Apple from Apple Business.
  • A strong Google profile is therefore not enough. The 2026 baseline: Google Business Profile, Bing Places, Apple Business, Yelp and the directories that matter in your country and industry, all with identical core data.
  • Reviews matter, but only genuine ones: Google bans incentives and review gating, and EU consumer law bans fake reviews.
  • Local landing pages help when they have real substance. City pages without their own content count as doorway abuse at Google.
  • Measure with a fixed set of 20 local prompts per month. Single answers are snapshots.

Local AI search means questions like "reliable electrician in Cologne-Ehrenfeld who can come this week?" being asked in ChatGPT, Gemini, Copilot or Perplexity instead of Google Maps. Instead of ten results, the AI usually names a handful of businesses. This guide explains where those names come from and what you can do about it, wherever in Europe you operate. Classic local SEO – ranking factors, profile guidelines, service areas – is covered in our guide to local SEO for businesses. Here we focus on the AI layer on top.

Why does AI search change how local businesses win customers?

Google's AI Mode has been available in Germany and the rest of the EU since 8 October 2025. ChatGPT shows maps with businesses for place-related questions on iOS and Android. For local businesses, the biggest change is the shortlist: an AI answer contains a few names rather than a long list of results.

The German directory publisher Gelbe Seiten measured this narrowing. According to its AI Visibility Report 2026 (9 September 2026, 10,000 searches across 10 industries in Germany), AI assistants named 4.8 businesses per search on average. 76.9% of businesses in Google's top 10 did not appear in AI answers, and almost 59% of the businesses named were not in Google's top 10. The publisher sells directory listings itself and describes its method only briefly, so treat the figures as an indication, not an independent finding. The direction still matches what practitioners see: a good Google ranking alone does not guarantee an AI recommendation.

Where do ChatGPT, Gemini, Copilot, Perplexity and Apple get local business data?

This is the key question for local businesses. The answer differs by system and is not fully documented everywhere.

Google: AI Overviews, AI Mode and Gemini

Google is the most explicit. According to Search Central, generative AI answers can include information about local businesses, and a maintained Google Business Profile helps you appear in AI answers and the rest of Search. In its list of basics, Google explicitly tells site owners to keep Merchant Center and Business Profile information up to date.

For places, the Gemini app uses public information from Google Maps – addresses, descriptions, websites, ratings and opening hours – according to Gemini Help. Your profile data does not come from you alone: Google combines crawled web content, licensed third-party data and user contributions, according to Business Profile Help. Conflicts between your website and your profile therefore get noticed.

New since June 2026: owners can connect their Business Profile to Gemini to analyse performance, draft review replies or update opening hours. That is a tool for you as the owner. Google does not say it improves visibility in Gemini's answers.

ChatGPT

OpenAI documents its sources only partly. According to OpenAI Help:

  • ChatGPT search sometimes partners with third-party search providers and rewrites your question into search queries such as "top restaurants San Francisco". As privacy notices, OpenAI links to Microsoft and Shopify. It does not literally say that Bing is the source.
  • For nearby business listings, ChatGPT shares approximate location with "trusted third-party providers", which it does not name.
  • For restaurant reservations, OpenAI names OpenTable (global), Resy (US) and Yelp (US and Canada).
  • At the launch of ChatGPT search, OpenAI said it had partnered with data providers for categories including maps.

On 23 July 2026, Yelp announced a licensing agreement with OpenAI: Yelp reviews, photos and business details appear in ChatGPT, with a "Request a Quote" feature to follow. The same report says Yelp already supplies Apple Maps, Microsoft Bing and Alexa+. For service providers, OpenAI is also testing a Get Quote integration with approved partners (beta). OpenAI has not published which providers it uses in which European country.

Many guides quote precise percentages, such as one data provider supplying most of ChatGPT's local results. We found no official confirmation; these figures come from one-off tests in single markets, so we leave them out.

Your own website reaches ChatGPT search through the OAI-SearchBot crawler. Sites that block it in robots.txt will not be shown in search answers, according to OpenAI.

Microsoft Copilot

Copilot is closely tied to Bing search: Bing Webmaster Tools counts Copilot citations in its AI Performance report. Microsoft's place for local business data is Bing Places for Business. According to Bing's help, it is free and controls listing details in Bing search. Bing checks submitted details against information from search engines, websites and social media, and recommends keeping core details consistent across all channels. You can import listings from Google Business Profile. Microsoft announced deeper integrations with Bing Maps and Copilot in October 2025 without describing them in detail.

Perplexity

Perplexity crawls with PerplexityBot to surface sites in its results (Perplexity docs). For restaurants it integrated Yelp data in 2024, and according to The Verge Tripadvisor followed for hotels in 2025. Perplexity does not publish a complete, current list of its local data sources.

Apple: Maps, Siri, Safari and Spotlight

Since 14 April 2026, Apple Business Connect has been part of Apple Business. According to Apple, you manage place cards with photos, opening hours and actions that appear across Apple Maps, Safari, Spotlight and more; the announcement also mentions Siri and Wallet. Brand and location management is available in all EU countries, the UK, Norway and Switzerland, according to Apple's availability list. Ads on Apple Maps are currently limited to the US and Canada.

Data-source matrix: which profiles feed which system?

AI system Local data according to the provider Other documented sources What you maintain
Google AI Overviews / AI Mode Business Profile, Merchant Center, Google index The profile itself draws on websites, licensed data, user contributions Google Business Profile, website, reviews
Gemini app public Google Maps data – Google Business Profile
ChatGPT third-party search providers, unnamed providers for business listings; OpenTable, Resy, Yelp for reservations Yelp licence (July 2026, per Yelp's announcement) Website (OAI-SearchBot), Yelp, Bing Places, booking platform
Microsoft Copilot Bing search; Bing Places for Bing search and maps Copilot citations in Bing Webmaster Tools Bing Places, Bing Webmaster Tools
Perplexity own crawler (PerplexityBot) Yelp (2024), Tripadvisor (2025) per media reports Website, Yelp, Tripadvisor (hospitality)
Apple (Maps, Siri, Spotlight) Apple Business Yelp lists Apple Maps as a partner Apple Business

As of November 2026. The last column is our inference from the sources, not a weighting of the providers.

NAP stands for name, address, phone number. For AI search, add website, opening hours, primary category, services and service area. The goal is that every source describes the same business – one clear entity that an AI system will not confuse with a namesake.

Step by step

  1. Define a reference record: write down the "truth" once: exact business name as registered or on your sign, address, main phone number, website URL, opening hours, one primary category, three to five core services.
  2. Check the core profiles: Google Business Profile, Bing Places, Apple Business, Yelp. Claim missing ones and align existing ones.
  3. Check national directories. These vary by country, for example Gelbe Seiten, Das Örtliche and 11880 in Germany, Herold in Austria, local.ch and search.ch in Switzerland, PagesJaunes in France, PagineGialle in Italy, Páginas Amarillas in Spain, De Telefoongids in the Netherlands and Yell in the UK.
  4. Check industry platforms (see the table below).
  5. Clean up duplicates and legacy data: old addresses after a move, former phone numbers, duplicate listings.
  6. Align your website: legal notice/imprint, contact page, footer and structured data must match the reference record exactly.
  7. Log it: for every source, record URL, status and date of the last check in a spreadsheet.

Industry platforms worth checking

Industry Platforms (examples; check which apply in your country)
Trades national tradesperson marketplaces (e.g. MyHammer in Germany), chamber of crafts directories
Medical practice doctor-booking and review platforms (e.g. Doctolib, jameda in Germany), official physician search services
Restaurant Tripadvisor, OpenTable, TheFork, Yelp
Real estate agent national property portals' agent directories, trade association member lists
Tax adviser / accountant official professional registers and chamber directories

The list shows where an audit is worthwhile. No provider discloses which of these sources an AI system actually uses, so we deliberately give no weighting. Paid premium listings are not needed for consistency; what matters is that the free listing is correct.

How do you collect reviews without breaking the rules?

Reviews are visible in almost every system: Google Maps, Yelp data in ChatGPT and Perplexity, Apple Maps. The rules are strict.

Google prohibits, according to its contribution policies:

  • incentives such as payments, discounts or free goods or services for posting a review or for changing or removing a negative one,
  • discouraging negative reviews or selectively asking only happy customers (review gating),
  • pressuring customers on site or requiring specific content,
  • setting staff quotas for reviews or for mentions of specific employees.

Asking for reviews of genuine experiences, without incentives and without influencing the content, is allowed.

Legally, the EU Omnibus Directive (EU) 2019/2161 added fake reviews to the blacklist of the Unfair Commercial Practices Directive: submitting or commissioning fake consumer reviews, or misrepresenting reviews, is always unfair. Businesses that display reviews must say whether and how they check that they come from real customers. Each member state has transposed this into national law, in Germany for example in the UWG. This is not legal advice; check individual cases with a lawyer. Regulated professions such as healthcare often have extra advertising rules.

What works in practice:

  • After every completed job, send all customers the same short link, without pre-selection.
  • Ask customers to describe what was done, without prescribing the wording. Specific reviews ("bathroom renovation in Ehrenfeld, on schedule") match specific AI questions better.
  • Reply to every review, factually and without personal data.
  • Make reviews possible on several platforms, not just Google.

Local landing pages: how do you avoid doorway spam?

Google's spam policies describe doorway abuse as, among other things, multiple pages or domains targeted at specific regions or cities that funnel users to one page, and substantially similar pages that resemble search results more than a clear site hierarchy. Thirty city pages where only the place name changes usually fall into this category.

A local page makes sense when it delivers its own answer:

  • real services and local specifics (e.g. older housing stock, listed buildings, typical jobs),
  • realistic price ranges and travel times,
  • named contacts and photos of real projects in the area,
  • answers to questions customers there actually ask.

Rule of thumb: if you could swap the place name without the page becoming wrong, it is too thin. One strong page with a service area beats ten weak ones.

Which structured data helps local businesses?

For LocalBusiness, Google requires name and address, according to Search Central. Recommended properties include geo, openingHoursSpecification, telephone, url and priceRange. Choose the right subtype, such as Electrician, Dentist or Restaurant. aggregateRating is only recommended for sites that capture reviews about other local businesses, not for your own stars.

Two points matter:

  • Markup must match the visible text. Google's AI features need no special schema, according to Google.
  • sameAs links your profiles (Google, Bing, Apple, Yelp, industry platform) to your website. It is a schema.org property that Google does not specifically recommend for LocalBusiness, but it can help other systems resolve your entity unambiguously.

20-prompt test protocol: how do you check your local AI visibility?

Prompt answers fluctuate heavily. SISTRIX found that ChatGPT replaces 74% of cited sources per week. Several different prompts measure much more accurately than one prompt repeated: with 20 prompts, the typical error is around ±9.5 percentage points, according to SISTRIX. That is enough for a trend, not for exact shares.

How to build your set of 20 prompts

Prompt type Count Pattern
Category + place 4 "[service] in [city/district]"
Recommendation 4 "Which [business] in [city] would you recommend?"
Condition 4 "… with [Saturday appointments / English-speaking / step-free access]"
Problem 3 "[specific problem] in [city], who can help fast?"
Comparison/criteria 2 "How do I recognise a good [business] in [city]?"
Price 1 "How much does [service] cost in [city]?"
Brand (control) 2 "What do you know about [business name] in [city]?"

Write prompts in the language your customers use. In multilingual markets such as Brussels, Zurich or Barcelona, split the set across languages and evaluate them separately.

Example prompts by industry (fictional)

Industry Recommendation Condition Problem
Trades "Which electrician in Cologne-Ehrenfeld is reliable?" "Electrician in Cologne with weekend emergency service" "Fuse keeps tripping in an old flat in Cologne – who can come today?"
Medical practice "Good dentist in Berlin-Friedrichshain?" "Dentist in Berlin with evening appointments, good with anxious patients" "Toothache at the weekend in Berlin, where do I go?"
Restaurant "Best vegetarian restaurant in Lisbon's Príncipe Real?" "Restaurant in Lisbon for 8 people with a booking tonight" "Where can I eat with a pram in central Lisbon?"
Real estate agent "Which estate agent in Vienna specialises in period apartments?" "Agent in Vienna who also values apartment buildings" "I've inherited a house near Vienna – who helps with the sale?"
Tax adviser "Accountant in Amsterdam for freelancers?" "Accountant in Amsterdam with digital bookkeeping and English-language advice" "Letter from the tax office, deadline in two weeks – accountant in Amsterdam?"

How to log results

  • Systems: ChatGPT, Gemini, Google AI Mode, Copilot, Perplexity. Each prompt in a fresh chat, logged out or in a private window.
  • Record context: date, system, device, location sharing (on/off), network. ChatGPT uses your approximate location from your IP address, according to OpenAI, so a VPN changes results.
  • Per answer: named (yes/no), position in the list, linked (yes/no), cited sources (e.g. Yelp, Google Maps, a national directory, your website), competitors named, details correct (yes/no, which field was wrong).
  • Analysis: mention rate per system as a monthly average; list of the most frequently cited sources. If a platform appears often, add it to your audit.

The "cited sources" column is the real value: it shows which directories AI systems actually use in your industry and city, regardless of what providers disclose. How to combine this with Search Console, Bing and GA4 is covered in how to measure AI search visibility.

Monthly routine: maintain local AI visibility in 60 minutes

  1. Profiles (15 min): check Google Business Profile, Bing Places, Apple Business and Yelp for user-suggested edits, opening hours (public holidays!) and new photos.
  2. Reviews (10 min): reply to new reviews on all platforms.
  3. Prompt set (20 min): run the 20 prompts in at least two systems and log them.
  4. Fix errors (10 min): correct wrong details from AI answers at the source – in the profile or directory, not in the chat.
  5. Decide (5 min): pick one action, e.g. add a missing industry platform or extend a service page with a frequently asked question.

Worked example: a painting business in Cologne (fictional)

This example is invented and not a client project. A painting and decorating business with one location in Cologne-Ehrenfeld notices that ChatGPT names three competitors for "painter in Cologne for an old apartment", but not them. Gemini names them, but with an old phone number.

  1. Log: the 20-prompt set repeatedly shows Yelp and a tradesperson platform as sources in ChatGPT and Perplexity. The business has no Yelp listing, and the platform shows the old address.
  2. Audit: the old phone number still appears in two directories and in an older field of the Google profile. Bing Places and Apple Business are unclaimed.
  3. Actions: define the reference record; fix the Google profile; create Bing Places via the Google import and check it; claim Apple Business; create a Yelp listing; correct the platform listing and directories; add LocalBusiness markup with sameAs.
  4. Website: instead of ten district pages, one strong page "period apartment renovation in Cologne" with service area, photos of real projects, price ranges and typical questions.
  5. Reviews: after every job, ask all customers for an honest review, alternating between Google and Yelp.
  6. Measure: the same set every month. Judge the trend after three months, not the first week.

Nobody can guarantee that ChatGPT will name the business afterwards. But these steps remove the provable reasons why it was missing or shown with wrong details.

Conclusion

Local AI visibility does not come from a single profile. Google uses the Business Profile, Copilot the Bing ecosystem, Apple its own place cards, and ChatGPT and Perplexity draw on third parties such as Yelp. If your data is identical and correct everywhere, your reviews are genuine and your local pages have substance, every system gets a clean basis. A fixed prompt set shows you which sources really count in your city.

The foundations for content and technology that AI systems can cite are in our GEO guide. If you want support: Contentoren's GEO service starts from €4,500 setup and SEO from €3,000 setup (both including German VAT). We clarify your starting point in a no-obligation call: get in touch.

Sources

Accessed 6 November 2026

Frequently asked questions

Where does ChatGPT get information about local businesses?

OpenAI says ChatGPT search sometimes partners with third-party search providers and shares approximate location with “trusted third-party providers” for nearby business listings, but it does not publish a full list. It names OpenTable, Resy and Yelp for restaurant reservations, and in July 2026 Yelp announced a licensing deal with OpenAI. Your own website reaches ChatGPT search through the OAI-SearchBot crawler if robots.txt allows it.

Is a well-maintained Google Business Profile enough for AI search?

For Google's AI Overviews, AI Mode and Gemini, the Business Profile is the key source of local details. ChatGPT, Copilot, Perplexity and Apple rely on other sources such as Bing Places, Yelp or Apple Business. If you only maintain your Google profile, the details in those systems are left to chance.

What is a NAP audit and why does it matter for AI recommendations?

NAP stands for name, address and phone number. An audit checks that these details, plus opening hours, website and category, are identical across your website, business profiles and directories. Microsoft explicitly says Bing compares Bing Places data with information from search engines, websites and social media, so conflicting data makes that check harder.

Can I offer customers a discount for a Google review?

No. Google prohibits incentives such as payments, discounts or free services for reviews, and also prohibits asking only happy customers. Across the EU, the Unfair Commercial Practices Directive as amended in 2019 bans submitting or commissioning fake consumer reviews. You may ask every customer equally for an honest review.

How do I test whether AI assistants recommend my business?

Build a fixed set of 20 prompts that mirror real customer questions in your city and run them monthly in ChatGPT, Gemini, Copilot and Perplexity, logged out, in a fresh chat, with the location noted. Record whether you are named, which competitors appear, which sources are cited and whether your details are correct. Judge trends, not single answers.