Google Business Profile (GBP) is the single highest-leverage AI visibility signal for a local service business. It’s the first authoritative source most LLMs read when constructing their representation of you, and it propagates into every directory that scrapes Google — meaning a single error on your GBP cascades across the web. Seven specific mistakes account for most of the AI-visibility damage we see: wrong primary category, mis-set service area, generic service descriptions, missing structured services, NAP inconsistency, stale review velocity, and an ignored Q&A section. Each is fixable in under 30 minutes.
If you read only one section of this post, make it the first — primary category is the foundation and the most commonly broken.
1. Wrong primary category
Your GBP primary category is the single most important field for AI citation. Models lean on it to decide what kind of business you are, full stop. If you’re a roofer but your primary category is “Construction Company” or “Home Improvement Contractor,” LLMs will not confidently surface you for “roofer” prompts — even if your business name contains the word “roofing.”
Concrete signs you’re affected:
- You appear in Google Maps results for generic queries (“contractor near me”) but not for your specific trade (“roofer near me”).
- Your competitors with the same trade have a category that’s narrower and more specific than yours.
- When you read your category aloud, it doesn’t match how customers describe what you do.
The fix: open your GBP → Edit profile → Business category → set Primary to the narrowest accurate term. “Roofing Contractor,” not “Construction Company.” “Plumber,” not “General Contractor.” “Endodontist,” not “Dentist,” if you primarily do endodontic work. Secondary categories let you cover adjacent services; the primary should be the single best description.
2. Service area mis-set
Two sub-mistakes here, both common:
Sub-mistake A: You have a storefront customers visit, but your GBP is set to “I deliver goods and services to my customers.” Result: your address is hidden, Google’s local pack treats you as a service-area business, and LLMs that rely on Google’s data lose your physical location entirely.
Sub-mistake B: You’re a service-area business (you go to customers), but your GBP shows a residential address as your storefront. Result: spam filters demote your listing, and your home appears on Google Maps — which is both a privacy problem and a trust signal failure.
The fix: GBP → Edit profile → Location and areas. If customers visit you, set “Yes” to “I have a storefront customers visit” and verify the address. If you go to customers, set “No” and add specific service areas (cities or zip codes — five to ten is the sweet spot; more than that gets filtered as keyword stuffing).
3. Service descriptions are generic or empty
GBP lets you add individual services (Plumbing Repair, Drain Cleaning, Water Heater Installation, etc.). Each one has a description field most businesses leave blank — or fill with one generic sentence.
These description fields are AI-citation gold. When ChatGPT or Perplexity browses Google to answer “who does tankless water heater installation in Phoenix,” the service description for “Tankless Water Heater Installation” on your GBP is one of the top passages the model retrieves. If yours is empty, you don’t compete for that prompt. If it says “Professional service you can trust,” you’ve handed the citation to whoever wrote something specific.
The fix: GBP → Edit services → for each major service, write a 50–80 word description that names the specific work, the service area, and any qualifying details. “We install tankless water heaters across the Phoenix metro, including under-warranty replacements on Navien, Rinnai, and Bosch units. Typical install: same-day for in-stock models, 48 hours for special order. Permits and inspection included.” That’s the kind of passage LLMs retrieve and cite.
4. Missing structured services
Adjacent to #3, but a separate field. Even if you’ve added a few services, most GBPs are missing services they actually do. Google offers an autocomplete list of services for every category — and the list is more comprehensive than most owners realize.
LLMs use the service list as a checklist. If you’re a plumber and your GBP lists only “Plumbing Repair” but the model knows your category typically includes “Water Heater Installation,” “Drain Cleaning,” “Sewer Line Repair,” etc. — and you don’t list those — the model won’t surface you for those specific prompts.
The fix: GBP → Edit services → review the suggested services for your category and add every one you actually perform. Aim for 8–15 services for most trades. Each one becomes a retrieval target.
5. Name, address, or phone inconsistency across the web
NAP consistency is older SEO doctrine, but it’s even more critical for AEO than it was for ranking. LLMs use cross-source matching to verify entity identity — if your business name resolves identically across Google, Yelp, BBB, Houzz, and your own website, the model has high confidence about who you are. If there are even small variations (“Smith Plumbing” vs “Smith Plumbing LLC” vs “Smith’s Plumbing of Phoenix”), the model treats these as possibly-different entities and recommends none of them.
Common drift points:
- Legal entity suffix (“LLC,” “Inc.”) in some places and not others.
- Phone numbers formatted differently — main line in some, tracking number in others.
- Suite or unit numbers that appear in some directories and not others.
- City spelled out vs. abbreviated (“Phoenix” vs “Phoenix, AZ”).
The fix: pick the exact canonical version of your name, address, and phone — including capitalization, punctuation, and formatting — and audit every directory listing. This is tedious but high-ROI. Start with the top 10 directories for your trade; that handles 80%+ of the corroboration weight.
6. Reviews dried up months ago
LLMs read your reviews — both the volume trajectory and the recency. A business with 30 reviews where the latest is from 18 months ago looks defunct or struggling. A business with 30 reviews where 8 are from the last 90 days looks alive and active. Models cite the second one preferentially, even at equal star ratings.
Two specific failure modes:
- Total stop: you stopped asking for reviews when your customer pipeline slowed, and now there’s a multi-month gap.
- Slow trickle: you get 1–2 reviews per quarter when your competitor is getting 1–2 per week.
The fix: build a systematic review request into your post-service workflow. Most service businesses can sustain 4–8 reviews per month with a simple ask — text or email within 48 hours of completion, with a direct link to your GBP review page. Not “could you leave a review when you have time” — specifically “would you take 60 seconds to leave a quick review? Here’s the direct link.” The conversion difference between those two requests is roughly 5x.
7. Q&A section is empty or filled with spam
GBP’s Q&A section lets anyone — owners, customers, random Google users — post questions and answers about a business. Most owners ignore it, which means it sits empty or fills with random questions left unanswered for years. Both states hurt.
LLMs read Q&A. When the section contains authoritative owner-answered Q&As, those become retrieval passages. When it contains stale unanswered questions (“Are you still open?” from 2022 with no response), the model treats it as a recency-failure signal.
The fix: seed your own Q&A with the 5–10 questions customers actually ask before booking — and answer them as the owner. Then, every quarter, scan for new questions and answer any unanswered ones. The 30-minute initial seed plus quarterly maintenance is one of the highest leverage GBP activities available, and almost nobody does it.
The 30-minute audit
Walk through your GBP with this list:
- Is the primary category the narrowest accurate term? (2 min to fix)
- Is the service-area setting correct for your model? (3 min)
- Does each major service have a 50–80 word description? (15 min for ~10 services)
- Are all the relevant services for your category added? (5 min)
- Are NAP details identical here vs. your top 5 other listings? (5 min to spot-check, more to fix everywhere)
- Is the most recent review within the last 30 days? (0 min to check, ongoing to fix)
- Are there unanswered Q&A entries? (2 min)
Most owners hit four or five of these in their initial pass. Each one closed is real AI citation share recovered — not theoretical, not marginal, and entirely free.