A multi-location local SEO audit checks every location in every town it serves on Google Maps, ChatGPT, Google AI Mode and Gemini, and totals the brand last. A brand total hides the misses. In the Cheers Market Baseline, Oct. 1 to 26, 2026, Google AI Mode and Gemini put the same business first for the same US question, city and week only 9.7% of the time, so a report's "Google AI" line averages two different shortlists.
Last verified: November 1, 2026.

The audit at a glance
If you only do one thing this week, run the first pass; it takes about an hour. Pick your weakest location and two towns it serves beyond the city on its address, and in a private window ask ChatGPT, Google AI Mode and Gemini one urgent and one planned question per town, naming the town and leaving out "near me". Log the 12 answers and a Google Maps search per town on the Answers tab, then repeat them the next day.
Run the five steps in order, since Step 3 checks what Step 2's answers turn up.
The audit, location by location
Fill every row before anyone reads the total
- 01
Map the towns
Every location and each town it sends trucks to, not only the city on its address.
- 02
Ask each engine
The same questions in each town on ChatGPT, Google AI Mode, Gemini and Google Maps.
- 03Evidence
Check the evidence
Profile, listings, local links, the phone number shown and the sources each answer cites.
- 04
Check reviews
Reviews in 90 days, days since the last one, and which profile the review link opens.
- 05Outcome
Add work and jobs
Work shipped that month, dated and linked, and booked jobs from the scheduling system.
Cheers, local SEO audit scorecard for multi-location brands, November 2026.
- Steps in a multi-location local SEO audit
- 1. Map the towns: Every location and each town it sends trucks to, not only the city on its address.
- 2. Ask each engine: The same questions in each town on ChatGPT, Google AI Mode, Gemini and Google Maps.
- 3. Check the evidence: Profile, listings, local links, the phone number shown and the sources each answer cites.
- 4. Check reviews: Reviews in 90 days, days since the last one, and which profile the review link opens.
- 5. Add work and jobs: Work shipped that month, dated and linked, and booked jobs from the scheduling system.
The scorecard: three tabs, three kinds of row
Reviews, profile checks and booked jobs belong to the location, not to each answer. Repeat booked jobs on every answer row at Step 2's depth (three towns, three questions, four surfaces) and you'd count them 36 times a month.
Download the spreadsheet or the Answers, Locations and Portfolio CSVs. Row 2 defines each column, and data entry starts in row 3. The workbook calculates the portfolio totals; the CSVs are manual templates.
Use the same brand and location names on every tab, with a unique location name within each brand. Enter Month as its first day. Pale yellow cells take your inputs and gray cells calculate. The workbook has room for 1,000 answers, 300 location-months and 100 portfolio rows; its notes explain what to extend for a larger audit. An unavailable check stays separate from a completed check that did not name the business.
Answers tab. One row per location, town, surface, question and date.
Its columns start with Brand and Location, as named on its profile; Town, the city on the address or a town it serves; Surface, which is ChatGPT, Google AI Mode, Gemini or Google Maps; Question, the exact wording with the town named and no "near me"; Date, the day you asked; and Named, in the answer or not (on Maps, in the top three or not). Then come First business (the first map card, or the first business named if there's no map), Town on the card (this town, another town, or no address), Phone shown (the number given for your location), Phone matches profile (whether it's the profile's number) and Top cited sources (the sites cited, in order), with the last three left blank on Maps rows. Save a link to the answer or screenshot in Evidence URL. The remaining columns calculate the reporting month and count each town once for that brand and surface, even if you tested it repeatedly or for several locations.
Locations tab. One row per location per month.
- Brand and Location: same names as on the Answers tab
- Month and Check date: the reporting month and the date you checked
- Reviews, last 90 days: new Google reviews in the 90 days before the check
- Last review date and Days since last review: enter the newest review date; the workbook calculates the gap
- Profile the review link opens: it should be this location's own
- Six profile checks: primary category, hours, services, booking link, a phone that reaches this location, a link to its page. Mark each Pass, Fail or Not checked; the total appears only when all six are checked
- Location page live: the page exists and the profile links to it
- Listings that match: Apple Business Connect, Bing Places, Yelp, BBB and cited directories showing the profile's name, address and phone
- Sites linking to the location page: local sites that link to it
- Work shipped (dated links) and Work items shipped: this month's changes, each dated and linked, plus their count
- Profile calls / directions / website clicks: from the Business Profile performance report
- Booked jobs: from the scheduling system
Our guide to those metrics explains Google's profile counts. Plan to copy them by hand, since Google's Sept. 15, 2026 notice on API applications said it "cannot provide a specific response timeframe" (Search Engine Roundtable).
Portfolio tab. One row per brand per surface per month.
Its columns are Brand, as customers know it; Surface, as on the Answers tab; Month, the month covered; Towns named at least once, meaning towns where any of the brand's locations was named on this surface that month; and Towns checked, every town on the Answers tab for this brand and surface that month. Four more come from the Locations tab: Locations with a 30-day review gap (Days since last review at 30 or more at this month's check), Locations passing all profile checks (6 of 6), Work items shipped (the dated items) and Booked jobs (the sum).
Towns come from completed Yes or No observations on the Answers tab. Location totals come once per location from the Locations tab and repeat on each surface row, so never add them across surfaces. The workbook flags duplicate location-months and missing inputs. Enter 0 when you checked and found none; leave unknown values blank. A 30-day review gap is a reason to investigate against that location's job volume and usual review pace, rather than a universal performance target.
Why one brand number hides the misses
Engines disagree (why AI recommends different businesses), and Google's two AI products are no exception. When both showed a map in those weeks, AI Mode and Gemini shared at least one business 51.9% of the time, against 58.2% for ChatGPT and Gemini.
First cards churn too. Across the US and Canada, AI Mode kept the same one the following week 23.5% of the time and Gemini, when it showed cards, 22.9%, so the Portfolio tab counts towns where a location was named at all.
Combine the answer checks with Google's reports. Business Profile performance measures profile interactions, while Search Console's Generative AI report reports AI-feature impressions that are also included in the Web total. The answer rows show the branch-specific recommendations those reports do not describe.
Cheers Market Baseline, US, October 1 to 26, 2026
Even Google's two AI products rarely put the same business first
Share of question, city and week combinations where both engines showed a map
Bars run from 0 to 100%.
ChatGPT and Gemini
2,934 combinations; 0.90 in common
2,934 combinations
Google AI Mode and Gemini
3,106 combinations; 0.73 in common
3,106 combinations
Cheers Market Baseline (our own research panel): 32 home-service questions asked weekly in 60 US metro areas, October 1 to 26, 2026. Queried November 1, 2026.
| Engine pair | At least one business in common | Same business in the first map card |
|---|---|---|
| ChatGPT and Gemini | 58.2% (2,934 combinations; 0.90 in common) | 9.9% (2,934 combinations) |
| Google AI Mode and Gemini | 51.9% (3,106 combinations; 0.73 in common) | 9.7% (3,106 combinations) |
Step 1: List every location and the towns it serves
The city on a location's address is only the first row of its territory, so add every town it sends trucks to and wants jobs from.
From Oct. 9 to 31, 2026, 18.4% of the map cards in US Google AI Mode answers and 14.9% in ChatGPT's named another town, so your branch, or a rival's, can get named one town over; the Town on the card column catches it.
In a big city, check whether you're on the dated best-of lists the answers cite; ChatGPT cited one in 54.0% of answers in the panel's 10 biggest US cities those weeks, against 29.9% in ten smaller ones.
Set the location the way our location-control protocol describes before you compare towns.
Step 2: Ask the same questions in every town, on every engine
Write three questions per trade per town: an urgent one, such as "who can fix a furnace today in [town]" for HVAC, a planned one and one about reviews and price. Name the town and skip "near me", which hands the engine your device's location, and ask in a private window, signed out where the engine allows it.
The first pass repeats a day later because "Answers can vary across runs, prompts, and time, making one-off observations unreliable" (Schulte, Bleeker and Kaufmann, arXiv, April 8, 2026). After it, keep the wording fixed; why AI recommends different businesses shows how far shortlists move week to week.
When Gemini shows no map, log the businesses it names in its text instead.
Maps needs its own rows, since "If a customer doesn't share where they are, Google uses what it knows about their location" (Google Business Profile Help), so ask a tech working in that town to screenshot the top three.
Keep Maps and AI Mode on separate rows, even though Google tested an "AI Overview" label on the local knowledge panel in September.
Ten locations, three towns each, three questions and four surfaces come to 360 checks a month, so size the job before anyone promises it.
Step 3: Check the evidence behind each location
Start with the profile. Google's guidelines say "Provide a phone number that connects to your individual business location, or provide a website that represents your individual business location" (Google Business Profile Help). Do both where you can. The guidelines also say "Do not create more than one page for each location of your business," so duplicates go on the fix list.
Then check each location's listings against its profile: Apple Business Connect, Bing Places, Yelp and BBB first, then the directories that town's answers cite. Our citation sources guide and Apple Business Connect for multi-location brands cover the details.
Note which local sites link to each location's page, since Google's ranking page counts "how many websites link to your business" toward prominence.
Next, check the number each answer printed. Of the US numbers ChatGPT printed in October 2026 that we could check, 17.6% weren't on the business's own website, and none of the confirmed nonmatches matched the single-digit or transposition errors we checked. Sites running call tracking hit 41.0%, against 12.1% without, so look for call-tracking swaps and old directory numbers before you change one.
Each town's answers cite different sources. Directories made up 21.8% of citations in Austin and 34.6% in Oakland in the four weeks to Sept. 2, so each town can need its own fix list.
If an acquired brand's old name keeps showing up, start with why AI treats your locations like strangers.
Next step
Is AI recommending your business?
Find out how visible you are across ChatGPT, Gemini, Perplexity, and AI Overviews.
Step 4: Reviews, location by location
At the typical home-service company with three or more trade locations connected to Cheers, the bottom half of its locations got 11.8% of its Google reviews from October 2025 to September 2026, where an even split would give them about 40%. Its busiest location got 41.5%. At the median company the busiest had 3.0 times the middle location's reviews, and 7 of the 11 companies were at twice or more.
Google reviews, October 2025 to September 2026
The bottom half of locations got about an eighth of the reviews
11.8%
of a typical company's Google reviews went to the bottom half of its locations
Base: median of 11 companies; an even split would give the bottom half about 40%
41.5%
of a typical company's reviews went to its single busiest location
median of 11 companies
3.0x
as many reviews at the busiest location as at the middle one
median of 11 companies
7 of 11
companies where the busiest location had at least twice the middle one's reviews
11 companies
Google reviews created October 1, 2025 to September 30, 2026 on Cheers-connected home-service profiles; deleted and replaced reviews excluded. Queried November 1, 2026.
| Finding | Value | Base |
|---|---|---|
| of a typical company's Google reviews went to the bottom half of its locations | 11.8% | median of 11 companies; an even split would give the bottom half about 40% |
| of a typical company's reviews went to its single busiest location | 41.5% | median of 11 companies |
| as many reviews at the busiest location as at the middle one | 3.0x | median of 11 companies |
| companies where the busiest location had at least twice the middle one's reviews | 7 of 11 | 11 companies |
Some of that is real demand, so compare reviews with completed jobs before you blame a location's asking process.
Quiet stretches are common: 36.4% of Cheers-connected trade locations with at least seven reviews in the year went 30 days or more without one. When a location hits that gap, check its asking process first.
Satellite reviews matter even when AI builds the shortlist. Of people who used AI for local business recommendations, 97% double-check its picks against reviews at least sometimes and 42% always check the review site itself (BrightLocal, March 10, 2026).
The fix is a process, so ask every customer at every location the same way. Google bars staff review quotas and says merchants shouldn't "selectively solicit positive reviews from customers" or "require or pressure users to leave ratings or write reviews while on the premises" (Maps user contributed content policy). Give every tech a simple neutral invitation and the right branch link, then use the attributed results to coach and recognize the team.
Open each location's review link yourself, because one shared link sends every location's reviews to one profile. See our review collection guide and why Google reviews go missing across multiple locations.
Step 5: Put work shipped and booked jobs beside the rows
Keep work shipped, the profile's calls, directions and clicks, and booked jobs in separate columns, so nobody reads a jump in clicks as a jump in jobs.
Don't judge October by GA4's AI referrals. Sistrix found ChatGPT tagged every source link with a utm_source tag set to chatgpt.com up to Oct. 8, 2026, then 1% to 24% a day from Oct. 9 to 15 and 42% on Oct. 16, and untagged visits land in Direct. Gemini dropped its utm_source=gemini tag around Sept. 28 and its iPhone app sends no referrer, Search Engine Roundtable reported Oct. 30, from one analyst's chart and no Google statement.
From a miss to a fix
- Not named in a town the location serves (marketing): add the town to the profile's service areas and location page, then re-ask (showing coverage to AI search).
- Named with a number that isn't on your site (marketing with the call center): match the page's default number to the profile, and keep old numbers ringing at that branch.
- A rival named, citing a directory where you're missing or listed differently (marketing): claim or correct that listing (citation sources; finding your industry's AI sources).
- A rival named and you can't see why (marketing): note what the cited pages say about the rival but not you (why AI recommends competitors).
- An old acquired brand name in the answer (marketing with the integration lead): find profiles, pages and listings still under the old name (why AI treats your locations like strangers).
- 30 days without a new review (location manager and operations): confirm the ask happens on every job and the link opens this profile (review collection).
- A profile check failing (location manager): fix the field and recheck next month.
Give every miss an owner and a date before the next monthly run.
The portfolio view for PE operating partners
Roll up only after the rows are filled, and count towns instead of averaging a score, since an average can't show which towns a brand is missing.
Franchise and roll-up brands compete location by location. In our US and Canadian panel, late August to September, national franchise brands appeared in 23.8% of answers naming a business but filled 7.5% of the named slots, which is why franchise locations win one at a time.
Use the rows to connect the work with changes in each market: what was corrected, how the branch appeared afterward and which enquiries became booked jobs. Dated source and answer records make that review useful.
For portfolio owners
Connect each market's visibility with delivery and booked work
The rows
It can show
- Which towns each location is named in, engine by engine
- Where a rival gets named and which sources the answer cites
- Profile, listing, page and phone gaps at each location
- Locations where reviews stopped arriving
- What shipped, when, and booked jobs beside it
Add the context
Read the changes accurately
- Connect dated fixes, repeated answer checks and booked jobs; a before-and-after change alone does not isolate the cause
- Mark engine changes, such as ChatGPT's October 8 to 9 layout switch, in month-over-month comparisons
- Label the October 9 to 16 ChatGPT referral gap, when most links lost their tag
- Combine referral reports with intake records for visits and calls that arrive without a source tag
Cheers analysis; Sistrix, October 30, 2026 (ChatGPT tags); Search Engine Roundtable, October 30, 2026 (Gemini tags). November 2026.
Looking at a target before close? Run the first pass on its location with the fewest Google reviews (you can count them from outside) and keep that dated row to compare after the deal.
The work-shipped column answers the question this ad opens with.
To hand the audit and the fixes to one team, see how we run multi-location local SEO.
Watch · 46 seconds
Who is making the changes this week?
For companies with 50 or more techs: one owner for the website, profiles and reviews, measured by market.
How often to run it
Run it monthly for every location and weekly for acquisitions, launches, seasonal peaks and your weakest markets, dating every row and keeping the old ones. When a row moves, compare its sources and format with last month's before anyone credits or blames the work; Sistrix found ChatGPT switched to a building-block layout overnight on Oct. 8 to 9, 2026.
Method
The Cheers Market Baseline, our own research panel, asks ChatGPT, Google AI Mode and Gemini the same 32 home-service questions every week, each naming the city, in 100 metro areas (60 US, 40 Canadian). We queried it Oct. 31 to Nov. 1, 2026.
Engine agreement uses US answers from Oct. 1 to 26, 2026, the weeks our panel captured Gemini map cards: 2,934 ChatGPT-Gemini and 3,106 AI Mode-Gemini combinations of question, city and week where both engines showed a map. "First" means the first map card; on first place the pairs were effectively tied (9.9% for ChatGPT and Gemini). We matched names after removing punctuation and case, so every agreement share is a floor. Week-over-week figures pool the US and Canada for the weeks of Oct. 5, 12 and 19 (5,355 week pairs for AI Mode, 1,640 for Gemini).
In the answers our panel captured, Gemini's US place-card share fell from 80.9% (Oct. 1 to 4, 837 answers) to 0.0% in the week of Oct. 26 (1,620 answers); we haven't confirmed whether Gemini changed or our capture did.
Card addresses cover 24,681 US AI Mode and 20,808 ChatGPT map cards, Oct. 9 to 31, 2026. A card named another town when its address didn't contain the city in the question, so suburbs and big-city neighborhoods with their own mailing names count. A dated list is a "best" or "top" page with 2025 or 2026 in its title or address; the ten smaller cities rank roughly 48th to 62nd in US population.
Directory shares (US median 26.9%) come from our published September data, the Home Services AI Visibility Index 2026: 60 US cities, four engines and 214,441 citations in the four weeks to Sept. 2, 2026. Franchise figures use 34,742 answers on four engines, Aug. 31 to Sept. 20, 2026. Phone figures use 900 numbers from ChatGPT answers, Oct. 2 to 31, 2026, checked against each business's own website on Nov. 1; 66 of 374 checkable US numbers weren't on the pages we checked.
Reviews: Google reviews created Oct. 1, 2025 to Sept. 30, 2026 on Cheers-connected profiles with a home-trade primary category and reviews before the window, minus deleted and replaced reviews. The within-company figures cover 11 companies with three or more trade locations, 100 locations and 25,529 reviews. A few large companies hold most of them, so we report the median company. "Bottom half" rounds down (1 of 3, 2 of 5), so an even split gives it about 40% at the median company and 46% pooled, where it got 12.2%. These companies aren't a market sample.
The drought figure covers 99 trade locations at 19 companies with at least seven reviews in the year, single-location companies and a few newer profiles included. Gaps before the first or after the last review don't count, so it's a floor. No business is named or identifiable.
Earlier versions quoted Cheers customer-panel rates; this one uses the Cheers Market Baseline.
Sources
- Google Business Profile API Application Processing Delays, Search Engine Roundtable, Sept. 15, 2026. Google's notice that it can't give a timeframe for API applications.
- Google Business Profiles Tests New Advanced Performance Reports, Search Engine Roundtable, Oct. 23, 2026. The tested report's sections, with no AI metric.
- AI features and your website, Google Search Central, updated Dec. 10, 2025. AI feature traffic counted as Web search.
- Don't Measure Once: Measuring Visibility in AI Search (GEO), Schulte, Bleeker and Kaufmann, arXiv, April 8, 2026. Why one-off checks are unreliable.
- Tips to improve your local ranking on Google, Google Business Profile Help, undated. How location and links shape local results.
- Google tests an AI Overview label on the local knowledge panel, Search Engine Roundtable, Sept. 17, 2026. The September label test.
- Guidelines for representing your business on Google, Google Business Profile Help, undated. Phone, website and one-page-per-location rules.
- Nearly Half of Consumers are Asking AI for Business Recommendations, BrightLocal, March 10, 2026. How often AI users double-check reviews; part of BrightLocal's Local Consumer Review Survey.
- Maps user contributed content policy, Google, undated. The Rating Manipulation limits on asking for reviews.
- What changed in ChatGPT in October: GPT-6, UTM tracking and sources, Sistrix, Oct. 30, 2026. ChatGPT's UTM tagging and layout change.
- Google Gemini Stopped Attaching UTM Parameters To Cited URLs, Search Engine Roundtable, Oct. 30, 2026. Gemini's dropped tag and missing iPhone referrer.
- Understand your Business Profile performance & insights, Google Business Profile Help, undated. What the current performance report counts.
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.
