AI Search Ranking Factors for Local Services in 2026 — What Actually Moves Mention Rates

Nobody published the algorithm — these are signal classes

OpenAI, Google, Anthropic, and Perplexity do not release local business ranking factor lists for their AI products. Anyone selling "the definitive AI algorithm" is guessing — or lying.

What we can document are signal classes that correlate with mention-rate movement when:

• Hundreds of buyer-intent prompts are sampled across platforms
• Competitor mention tables are compared to verifiable public data
• Signal fixes are applied and resampled monthly

This article organizes those signal classes for local service businesses — plumbers, HVAC, dentists, lawyers, med spas, electricians, and similar high-intent categories — with honest caveats, platform nuance, and measurement guidance.

No factor here guarantees placement. Together they describe what actually moves mention rates when executed consistently.

Related: AI visibility tracking, AEO, GEO.

Research context: Scan invisibility study — anonymized free-scan aggregates when sample size permits.

How to read this article

Each factor includes:

• What it is — operational definition
• Why models appear to weight it — retrieval and synthesis logic
• Local service examples — vertical specificity
• How to audit — practical checklist
• Caveats — where overconfidence fails

Factors are ordered by leverage for typical SMB starting points, not by a secret official weight formula.

Strategy foundation: How AI assistants choose businesses.

Factor 1 — Review volume, rating, and recency

What it is

Aggregate star rating, total review count, velocity of new reviews, and distribution over time on public review surfaces — primarily Google, plus Yelp and vertical sites depending on category.

Why models weight it

Reviews bundle social proof in natural language models parse easily. Star counts and review totals are groundable facts. Review text supplies themes models map to user intent.

Local examples

• Plumber: reviews mentioning "same-day," "cleaned up," "explained pricing"
• Dentist: reviews mentioning "gentle," "insurance," "short wait"
• Lawyer: reviews mentioning "responsive," "clear communication," "outcome" (within bar rules)

How to audit

Check	Pass?
Review count within competitive range for city
Rating at or above category norm
New reviews in last 90 days
Theme keywords matching top buyer prompts
Owner responses on recent reviews
No gating policy violations

Ethical playbook: Google reviews the right way.

Caveats

• Threshold effects — Below competitive review count, other factors may not matter yet
• Rating alone — 5.0 with 8 reviews loses to 4.7 with 300 on competitive prompts
• Fake reviews — Platform penalties and model distrust; never worth it

Factor 2 — Review theme alignment with buyer prompts

What it is

Semantic overlap between how buyers ask ("emergency AC repair," "veneers not braces," "custody modification lawyer") and recurring phrases in review text and service labels.

Why models weight it

AI recommendation is intent matching under uncertainty. Themes reduce uncertainty — if forty reviews mention "weekend hours," the model infers weekend availability more confidently than a generic 4.9 rating.

Local examples

• HVAC: "came same day during heat wave" aligns with emergency prompts
• Orthodontist: "Invisalign" in reviews aligns with cosmetic teen prompts
• Criminal defense: "DUIs" and "court communication" align with practice-area queries

How to audit

1. List your top ten buyer-intent prompts
2. Search your Google reviews for theme keywords
3. Compare to competitors named most in AI scans
4. Gap analysis → service delivery coaching + ethical review requests

Caveats

• Scripted fake theme stuffing in reviews violates policies
• Themes must reflect real operations — mismatch creates bad jobs and accuracy repair work

Factor 3 — NAP consistency and entity resolution

What it is

Name, Address, Phone — plus website URL and suite numbers — matching across Google Business Profile, Apple Business Connect, Bing Places, Yelp, and vertical directories.

Why models weight it

Entity resolution systems merge or split business records. Inconsistent data creates ambiguous entities models skip rather than risk wrong recommendations.

Local examples

• Two listings with different suite numbers → competitor gets the mention
• Old call tracking number on Yelp, new number on GBP → model shows wrong phone
• DBA versus legal entity name drift → split review corpus

How to audit

Export every live listing to a spreadsheet. Flag character-level mismatches. Priority-fix Google + Apple + top revenue-driving directories.

Guide: Apple Business Connect.

Caveats

• Aggregator sites you do not control may lag — monitor, don't ignore
• Duplicate GBP listings require Google merge process, not schema alone

Factor 4 — Google Business Profile completeness and freshness

What it is

Services list depth, accurate hours, attributes, photos, posts, Q&A, service area definition, and review response activity on GBP.

Why models weight it

Gemini and Google AI Overviews lean on Google's local graph. GBP is the operational truth node for Google-ecosystem AI — and corroboration for other engines scraping public data.

Local examples

• Electrician listing "panel upgrade" and "EV charger install" separately — not just "electrical services"
• Med spa with treatment-specific services and policy-compliant photos
• Law firm with practice areas listed individually

How to audit

GBP element	Status
All services listed
Hours including holidays
Service area accurate
10+ photos, refreshed this quarter
Post in last 14 days (competitive markets)
Q&A seeded with real questions

Compare: ChatGPT vs Google AI Overviews.

Caveats

• GBP spam categories hurt trust — list what you actually do
• Automated posting without approval creates brand risk

Factor 5 — Cross-directory and vertical listing presence

What it is

Presence and accuracy on category-specific directories beyond Google — Angi, HomeAdvisor, Healthgrades, Zocdoc, Avvo, BBB where appropriate.

Why models weight it

Retrieval-forward engines (Perplexity, Claude browse modes) and ChatGPT browsing pull multi-source corroboration. Businesses named on three directory types beat single-source dependence.

Guide: How Perplexity cites local businesses.

Local examples

• Contractor on Angi + GBP + BBB with matching NAP
• Dentist on Healthgrades + Zocdoc + GBP
• Lawyer on Avvo + state bar directory + GBP

How to audit

Identify top three directories buyers use in your vertical locally. Claim, verify, sync.

Caveats

• Pay-to-play lead gen directories with fake profiles — fix or remove
• Incomplete profiles worse than unclaimed in some cases — half-truths propagate

Factor 6 — Entity schema, llms.txt, and crawlable facts

What it is

JSON-LD LocalBusiness markup, llms.txt summary, fact-dense service and FAQ pages, sensible robots.txt allowing public crawl.

Why models weight it

Structured data reduces parser ambiguity. Citable sentences on crawlable URLs give retrieval systems quotable anchors.

Technical checklist: llms.txt, schema, and robots.

Local examples

• areaServed cities matching real service area
• hasOfferCatalog listing actual services with descriptions
• FAQ: "Do you offer financing?" with specific answer

How to audit

• Rich Results Test for schema validity
• yoursite.com/llms.txt exists and matches live facts
• robots.txt not blocking service pages

Caveats

• Schema alone rarely moves competitive mention rates without reviews
• Blocking all AI crawlers may reduce grounding accuracy — strategic choice, not default advice

Factor 7 — Citable third-party content

What it is

Pages and URLs models can quote — data studies from public datasets, merit-based press, authoritative listicles with editorial standards, deep FAQ with sourced claims.

Why models weight it

Generative synthesis prefers grounded claims. "We are the best" fails; "According to county permit data…" cites when methodology is on-page.

Local examples

• Roofer publishing hail claim frequency by ZIP from public NOAA data
• Law firm publishing statute deadline FAQ with citations
• Dentist publishing transparent pricing ranges where legal

How to audit

Does your site have one URL worth citing for your top buyer prompt? If no, add to content queue.

Caveats

• Generic AI blog spam adds little
• Press without verifiable hook gets ignored

Factor 8 — Third-party mentions and legitimate recognition

What it is

News coverage of real milestones, merit-based awards, community sponsorships with independent reporting — not purchased badge graphics.

Why models weight it

Independent mentions corroborate entity prominence beyond self-published claims.

Caveats

• Pay-to-play "best of" badges with no editorial bar — low durable signal
• Fake press releases damage credibility when models cross-check

Factor 9 — Geographic and service fit

What it is

Clear service area boundaries in listings, schema, and site copy — matching user geo intent in prompts.

Why models weight it

Recommending a business outside service area fails users. Models infer fit from areaServed, GBP service area, and review locations.

Local examples

• "Near me" prompts favor businesses with tight geo relevance signals
• Multi-location brands need location pages with distinct NAP — not one blended profile

Caveats

• Keyword-stuffing city names in footer spam — ignore or penalize
• Fake satellite offices — entity confusion risk

Factor 10 — Recency and operational signals

What it is

Recent reviews, fresh GBP posts, updated hours, active Q&A responses — signals the business is operating now.

Why models weight it

Stale profiles lose urgency prompts. Competitors with last-week reviews beat dormant listings.

How to audit

Last review date, last GBP post, hours updated for season/holidays.

Caveats

• Burst fake recency triggers spam classifiers
• Steady ethical velocity beats spikes

Factor 11 — Factual accuracy across the corpus

What it is

Consistency of hours, phone, services, and status (open/closed) everywhere AI might read — including outdated press and wrong Yelp hours.

Why models weight it

Wrong facts do not always zero mention — sometimes they replace you with a competitor when your data looks unreliable.

Repair guide: AI reputation repair.

Caveats

• Fixing the model directly does not work — fix sources
• Resample four to six weeks after corrections

Factor 12 — Platform-specific source mixes (measurement factor)

What it is

Not a single input — recognition that ChatGPT, Gemini, Claude, Perplexity, and Grok weight overlapping but distinct source sets.

Why it matters for ranking factors

Improving Google-only signals moves AI Overviews and Gemini more than ChatGPT-only blind spots — and vice versa for cross-web review density.

Overlap data: ~11% shared domains in some cross-platform samples — methodology varies.

Read: Eleven percent problem.

How to audit

Platform-by-platform mention table from free scan. Do not aggregate into one false "AI rank."

Factors that matter less than marketers claim

Claimed factor	Reality for local services
Keyword density on homepage	Weak vs review themes
Backlink count alone	Moderate — citations help, not sufficient
Social media follower count	Weak direct signal in sampling
Paid AI placement	No legitimate organic option
Single schema deploy	Necessary, rarely sufficient
Blocking AI crawlers	Often hurts accuracy

Traditional SEO still matters for clicks and site conversion — it just does not fully predict AI mentions.

Compare: AEO vs GEO vs SEO.

Weighting varies — composite scenarios

Scenario 1 — New business, zero reviews

Bottleneck: Factor 1 threshold — without reviews, other fixes rarely move competitive prompts.

Priority: Ethical review velocity + perfect listings + schema. Expect months, not days.

Scenario 2 — 200 reviews, wrong AI hours

Bottleneck: Factor 11 accuracy — listing conflict undermines trust.

Priority: Accuracy repair before new content campaigns.

Scenario 3 — Strong Google, invisible on ChatGPT

Bottleneck: Factor 12 platform mix — cross-web corroboration thin.

Priority: Vertical directories + citable site FAQ + Yelp themes.

Playbook: LLM SEO for local business.

Scenario 4 — Mentioned but competitor wins share

Bottleneck: Factor 2 theme alignment or Factor 1 volume gap.

Priority: Theme coaching in service delivery + review requests; compare competitor review text honestly.

Measuring factor impact — scientific enough for business

1. Baseline mention rate on fixed prompt set — six-platform scan
2. Hypothesis — which factor class is weakest vs named competitors
3. Intervention — fix one class seriously for 60–90 days
4. Resample — same prompts, same platforms
5. Attribute — mention delta + calls booked

Change multiple factors at once without measurement and you learn nothing.

AI visibility tracking is the discipline — not one-time audits.

Invisibility baseline

Anonymized scan aggregates suggest a meaningful share of local businesses are never mentioned on sampled buyer-intent prompts — exact percentages published only when sample size exceeds thresholds in the invisibility study.

If your baseline is zero mentions, factor 1 and 3 usually precede factor 7.

Execution priority matrix

Starting condition	Fix order
Listing conflicts	Factor 3 → 11
Thin GBP	Factor 4
Low review count	Factor 1 → 2
Accurate but invisible	Factor 5 → 7 → 8
Platform-specific gap	Factor 12 targeted
Wrong AI facts	Factor 11 immediately

GEO guide maps factors to Growth Engine modules.

What honest programs refuse to promise

• Guaranteed #1 AI placement on any factor combo
• Instant mention spikes from schema-only projects
• Identical rankings across ChatGPT and Google AI Overviews
• "Secret" factors requiring proprietary payment to OpenAI or Google

Factors move odds. Measurement proves movement — or proves you need a different intervention.

Next steps

AI search ranking factors for local services in 2026 reduce to verifiable public evidence:

1. Reviews with prompt-aligned themes

2. Consistent listings across directories

3. Complete, fresh GBP

4. Entity schema and crawlable facts

5. Citable content and accurate corpus-wide data

6. Monthly multi-platform measurement

7. Scan your visibility — establish mention-rate baseline

8. Audit factors against top three named competitors

9. Execute highest-leverage class for 60–90 days

10. Resample; attribute booked jobs, not just mentions

Services: AEO, GEO, LLM SEO. Tracking: AI visibility tracking.

The algorithm stays secret. The signal classes do not. Fix what models can verify — and measure whether mention rates move.