Skip to main content
Cheers

AI Search News

How Does Meta Muse Find Local Businesses?

Muse browses the web and acts for its user. Prepare your service information, reputation and booking paths for an assistant researching local providers.

Dylan Allen-Arnegard, CEO and Co-Founder of Cheers
Dylan Allen-Arnegård

CEO & Co-Founder, Cheers

11 min readPublished

Last verified .

A grey-haired man in a tan canvas coat, seen from behind on an empty commercial street, tries the chained, papered-over door of a plumbing shop that closed years ago, with weeds growing along its threshold

Meta Muse gives a customer an assistant that can open websites, fill forms and act on their behalf. Meta's September 8, 2026 launch post describes its own browser, connected apps and memory. For a local service business, prepare the information and contact paths that assistant needs: clear service pages, a verifiable reputation, accurate listings and a working way to request the job.

Last verified: November 1, 2026.

We expect personal agents to take on more of the buying journey. The work starts with making your business easy to evaluate and contact through the tools they already use.

What Meta's own pages say about how Muse looks things up

Meta's launch post puts Muse on what it calls Muse Secure VM, a dedicated computer with its own browser, where Muse can open pages, fill out forms and negotiate for the person using it. That person picks which apps Muse connects to. Muse also remembers what matters to them, and they can tell it to forget specific things it has learned.

Those capabilities let Muse work across public pages and the services its user connects. Meta has not published a fixed source list for local recommendations, so use direct checks to see which pages it consults for your trade and city.

Connectors give Meta's AI a direct route to outside services. The developer connectors page is a preview whose examples are fitness, productivity, entertainment and smart home, and the eight early local partners in the Singapore launch post on October 27 span shopping, groceries, dining and travel, with no home service among them. TechCrunch reported on October 6 that Meta also keeps a growing connector list inside the Muse app.

On the record, as of Nov. 1, 2026

Prepare the pages Muse can use and check the connections it supports

On the record

What Meta's pages and one reporter's test show

  • Muse can open a browser, fill out forms and negotiate (Meta launch post, Sept. 8, 2026)
  • People choose which apps Muse connects to (launch post)
  • Muse remembers what matters to the person using it, who can tell it to forget things (launch post)
  • Meta-WebIndexer crawls the web for Meta AI search, and allowing it helps Meta cite and link a site (Meta crawler page)
  • Meta-ExternalFetcher fetches links at a user's request (Meta crawler page)
  • Muse said it found no mention of an item on a restaurant's own site and suggested calling (Yahoo Finance reporter's test, Sept. 23)

Questions for your provider

Check the path in your market

  • Which pages Muse opens when you ask for your trade in your town
  • Whether the sources it uses carry the right branch, services and phone number
  • Whether your quote form reaches the branch and receives a useful response
  • Whether your firewall lets a legitimate customer-directed browser complete the request
  • Which Muse connectors your scheduling provider supports, and what each one can do
Meta documents browser tasks and app connections. The right column is our practical check for a branch; inspect the actual request path as the product develops.

Meta newsroom (Sept. 8, 2026); Meta crawler documentation (undated); Yahoo Finance, Daniel Howley (Sept. 23, 2026); TechCrunch (Oct. 6, 2026). Meta pages read Nov. 1, 2026.

What Muse did when a reporter tried it on local businesses

Daniel Howley's September 23 review for Yahoo Finance shows Muse handling local requests for movers and restaurants from one user's account.

With movers, Muse didn't stop at a list. It took his current and new addresses, his number of bedrooms and his contact details, then found four moving companies in his area and sent out his request for quotes; about two minutes later, two of them were calling his phone. That is the customer journey to prepare for: an assistant can turn a request into conversations with several providers.

Later he asked for nearby restaurants that serve gluten-free beer. Two of Muse's three picks were right, but the third had been closed for years, and when he pointed that out, Muse said it had relied on stale data at first.

Then he asked about a spot he already knew carried it, and Muse said it couldn't find that beer mentioned on the restaurant's own site and suggested he call. Give that assistant a clear answer on your own service page.

Muse also gave him a phone number, then told him to use a different one, and when he asked why, it said it had made up the first number without confirming it.

Meta's crawlers, and what local websites do about them

Meta's undated web crawlers page is the only Meta page we found that explains how Meta's AI reads websites. Meta-WebIndexer crawls the web to improve Meta AI's search results, and Meta says allowing it in robots.txt "helps us cite and link to your content in Meta AI's responses." Meta-ExternalFetcher fetches links a user asks for, which Meta says it uses to evaluate and improve its agentic AI, and it may bypass robots.txt rules. Meta-ExternalAgent crawls to train foundation models and to index content for Meta's products.

The page says Meta AI, not Muse. Meta hasn't said whether Muse's browser starts from Meta-WebIndexer's index, Google, Bing or something else, and nothing we read says whether that browser follows robots.txt.

Few contractor sites block AI bots at all. When Cheers checked 947 US home-service contractor sites with a valid robots.txt on Nov. 1, 2026, 6.1% blocked at least one AI bot and 1.6% blocked an AI search or user agent. In a second check, 1.4% blocked Meta-WebIndexer, and every one of those also shut out Googlebot, so none was singling out Meta.

US contractor websites, Nov. 1, 2026

Few contractor sites block AI bots at all

6.1%

block at least one AI bot

Base: 58 of 947 robots.txt files

1.6%

block an AI search or user agent

15 of 947

Meta's crawler page doesn't say Muse uses Meta-WebIndexer's index, and Meta hasn't said whether Muse follows robots.txt. DataDome, which sells bot protection, found Muse's browser didn't identify itself as an AI agent from Sept. 22 to Oct. 3. PPC Land reported on Oct. 31 that a DataDome researcher has seen Meta signing a small number of requests from Muse and Meta-WebIndexer since about Oct. 12; Meta hasn't confirmed it.

Cheers check of 1,096 US home-service contractor websites from Google's local map results (Cheers customers removed); 947 valid robots.txt files, Nov. 1, 2026. DataDome via PPC Land (Oct. 15 and Oct. 31, 2026).

6.1% of 947 robots.txt files on US home-service contractor websites block at least one AI bot. 1.6%, or 15 files, block an AI search or user agent.
FindingValueBase
block at least one AI bot6.1%58 of 947 robots.txt files
block an AI search or user agent1.6%15 of 947

Blocking a crawler by name only works on a bot that gives its name. DataDome, which sells bot protection and doesn't say how big its network is, found that Muse's browser didn't say it was an AI agent in the requests it counted from September 22 to October 3, PPC Land reported on October 15.

That may be changing. On October 31, PPC Land reported that Jérôme Segura, DataDome's VP of Threat Research, wrote on LinkedIn that Meta seems to be quietly testing Web Bot Auth, which signs requests so a site can check who sent them, for at least Meta-WebIndexer and Muse since around October 12. He said DataDome still sees only a relatively small number of signed Meta requests. Meta hasn't confirmed any of this, and its crawler page publishes nothing a site could use to check those signatures.

Meta, a founding member of the Overture Maps Foundation, builds Overture's places data into the maps on Facebook and Instagram, Overture says, but nothing we read ties that data to Muse.

Build the sources that make your business easy to choose

When I explain this to owners, I focus on three places: your own website, review and directory profiles such as Google, Yelp and the BBB, and independent mentions such as a Chamber of Commerce page. Together, those sources give a customer or assistant a way to understand the business and check its reputation.

Muse adds the person's connected apps and remembered preferences to the process. That makes strong public information especially useful: the assistant needs to work out whether your services fit this particular customer. Publish the jobs you do, where you work, your credentials and the next step in plain language.

What to prepare now

Make agent readiness part of the branch's visibility plan. Keep the site accessible, make the records agree and check Muse directly in your main markets.

Start with robots.txt this week. Open yoursite.com/robots.txt and look for a "Disallow: /" line under "User-agent: Meta-WebIndexer" or under "User-agent: *", the rule that shuts out every crawler; if either is there, ask whoever runs your site to take it out. Meta-ExternalAgent, the training crawler, is a separate call that belongs with your choices on OpenAI's and Anthropic's bots, and our guide to which AI crawlers to allow walks through them one bot at a time.

Every couple of weeks, ask Muse on your phone the question your customers ask, for two towns you serve: who's the best [trade] in [town] for [job]? Muse tailors its answers to what it remembers about the person asking, so also have a friend who has never talked to Muse about your business ask, or tell Muse to forget what it has learned about your company before you start.

Ask for names only. If Muse offers to call, message, request quotes or book, say no, and don't give it your phone number, email or address, because Howley's test showed how fast a request puts real companies on the phone. Screenshot whatever it names, and don't hand your Meta login to any AI tool that offers to run these checks for you.

Check each engine directly. The engines we track choose substantially different businesses: when ChatGPT and Google AI Mode both showed businesses on a map for the same question, US city and week (Oct. 12 to 31, 2026), they had 0.91 businesses in common out of about four each, and in 41.7% of those cases none at all. Muse browses on its own and remembers each user, so there's no reason to assume its list copies either one.

For the full split of which businesses each engine names, see why AI recommends different businesses every time. Part of the reason two engines disagree is that they read different sources, as how AI search engines differ in the sources they cite shows.

This month, put the facts a customer's agent would be asked for in plain text on pages anyone can read without signing in: your service area, hours, services, a starting price or range, how to book and a phone number. Meta says Muse fills out forms in its own browser, so the form checks in can AI agents book appointments from your website apply to it too.

Your own site is the one source you control, and it's the one Muse checked in the Yahoo Finance test. Other places limit what a bot or a signed-out visitor can read: Yelp told TechCrunch it doesn't permit non-human traffic unless the agent pays through its data licensing program, and Search Engine Roundtable reported on September 30 that Google Maps now asks people to sign in before they can read, sort or search all of a business's reviews.

Also this month, make your name, address, phone number and hours match on your website, Google Business Profile and Facebook Page. Consistent records give the assistant a clear, current identity to work with. If ChatGPT or another AI still repeats a closed location or an old number, start with how to fix wrong business information in ChatGPT.

And if someone sells you a Muse ranking, ask how many Muse answers they logged, from which account and in which cities. A score needs that evidence before it belongs in your plan; our source review did not establish a public benchmark.

Method

This is a source-based guide to Muse and an operational recommendation. The original Cheers data below covers contractor websites and other AI engines.

  • Meta's pages: the Muse launch post (Sept. 8, 2026), the Singapore launch post (Oct. 27, 2026), the web crawlers page (undated), the connectors page (undated) and the Connect recap (Sept. 24, 2026), all read on November 1, 2026. The reporter's examples come from Daniel Howley's Sept. 23 Yahoo Finance test, run from one account; they aren't a home-service benchmark.
  • The Cheers Market Baseline is our own research panel: every week it asks ChatGPT, Google AI Mode and Gemini the same 32 home-service questions in 100 metro areas, 60 in the US and 40 in Canada. This page uses the 60 US cities, ChatGPT and Google AI Mode, and the weeks of Oct. 12, 19 and 26 (Oct. 12 to 31, 2026), starting with the first full week after ChatGPT changed its local answers on October 9. The averages of about four (4.21 for ChatGPT, 4.26 for AI Mode) count each engine's question-city-weeks with map cards. The overlap uses only the 4,125 question-city-weeks in which both showed a map; of all 5,457 US question-city-weeks, ChatGPT showed one in 81% and AI Mode in 93%. We match names after removing punctuation and case, so spelling variants can hide a match and the overlaps are floors.
  • Robots.txt: On Nov. 1, 2026, Cheers read the homepages of 1,096 US home-service contractor websites drawn from Google's local map results, with Cheers customers removed; 976 were readable as served. 947 returned a valid robots.txt file, and we checked each crawler's access to the homepage: 58 blocked at least one AI bot and 15 blocked an AI search or user agent. The Meta-WebIndexer rate comes from a second check the same morning, in which 904 sites returned a valid file and 13 blocked Meta-WebIndexer.
  • November 1, 2026: first version.

Sources

Sources read November 1, 2026.

Dylan Allen-Arnegård is the CEO & Co-Founder of Cheers, the done-for-you local visibility partner for established home-service companies and PE-backed portfolios.

Frequently Asked Questions

Meta hasn't said. Its launch post says Muse works in its own browser, and its crawler page describes Meta-WebIndexer, which feeds Meta AI search, but no Meta page we read on November 1, 2026, names the search engine or index Muse's browser starts from.

None of the Meta pages we read says so. Muse for Small Business, launched September 29, connects an owner's own Facebook Pages, Instagram analytics and ad accounts; that says nothing about what a customer's Muse reads. Overture says Meta uses its places data in Facebook and Instagram maps, so keep your Page's address, hours and phone current, though Meta hasn't tied that data to Muse.

Muse can browse business websites. In Yahoo Finance's September test, it checked a restaurant's site for a specific item. Publish services, areas, hours, credentials and booking instructions as accessible text so it can find the details needed for a customer's request.

Make the business easy to find, evaluate and contact: keep the website accessible, publish specific service and location information, maintain the review and directory pages that establish your reputation, and test Muse's answers in your markets. Meta has not published a local ranking formula; these are the practical foundations we recommend for serving customers through personal agents.

Keep Meta-WebIndexer open, since Meta says it helps Meta AI cite and link your site. Few contractor sites block AI bots at all: of 947 US contractor sites with a valid robots.txt that we read on Nov. 1, 2026, 6.1% blocked at least one. Meta-ExternalAgent, the training crawler, is a separate choice, and Meta hasn't said whether Muse's browser follows robots.txt.

Written by

Dylan Allen-Arnegard, CEO and Co-Founder of Cheers

Dylan Allen-Arnegård

CEO & Co-Founder, Cheers

Dylan leads Cheers, the done-for-you local visibility partner for established home-service companies and PE-backed portfolios. He started out helping frontline teams earn reviews.

Share this article

Muse browses the web and acts for its user. Prepare your service information, reputation and booking paths for an assistant researching local providers.

Next step

Is AI recommending your business?

Find out how visible you are across ChatGPT, Gemini, Perplexity, and AI Overviews.

Keep reading