Quarterly AI Visibility Audit Checklist for Local Business

Why quarterly — and why "set and forget" fails

A med-spa owner ran our free scan in March. Zero mentions on four platforms. She fixed Apple Business Connect, cleaned a Yelp duplicate, and sampled monthly through summer. Mention rate hit 35% on Gemini by August.

She skipped September and October — busy season. November rescan: 12% on Gemini, competitor up 40%. A new rival had run an aggressive review campaign and published FAQ content on "Korean vs European skincare protocols" — themes Gemini started echoing.

Nothing " broke." The market moved and she was not watching.

Quarterly AI visibility audits are not bureaucracy. They are the minimum cadence to detect mention drift, listing decay, wrong facts, and competitor SOAV gains before they show up as unexplained lead softness.

This checklist is the operational doc we give Growth and Dominance clients at AIrecommend.ai — adapted for owners and marketing leads who DIY all or part of AEO. It complements how to check what ChatGPT says with a full-funnel review: measurement, accuracy, entity, content, competition, attribution.

Honest limit upfront: no audit guarantees placement in third-party AI products. Audits guarantee visibility into whether your signal work is working — which is the prerequisite for intelligent fixes.

Audit architecture — six workstreams

Each quarter, run these blocks in order. Do not jump to content creation before accuracy passes.

Quarterly AI Visibility Audit
├── 1. Measurement (mention + SOAV baselines)
├── 2. Accuracy (NAP, hours, services, AI wrong facts)
├── 3. Entity (schema, llms.txt, directory graph)
├── 4. Reviews (velocity, themes, responses)
├── 5. Competition (SOAV delta, citation gaps)
└── 6. Attribution + planning (calls, next-quarter priorities)

Estimated time:

Locations	DIY hours/quarter
1	3–5
2–5	6–12
6+	Consider AIrecommend.ai Growth/Dominance or dedicated ops

Schedule audits same week each quarter — e.g., second Monday of Mar/Jun/Sep/Dec — to avoid holiday distortion and align with fiscal planning.

Workstream 1 — Measurement baseline

1.1 Confirm prompt library (30 min)

Your library should include 5–15 buyer-intent prompts per location — category + intent + geography. No brand-leading questions except reputation repair tests.

Examples:

• "Best emergency electrician in [city] open now"
• "Who should I hire for a kitchen remodel under $80k in [county]?"
• "Recommend a trustworthy CPA for small business near [neighborhood]"

Quarterly task: Archive prompts retired this quarter; add prompts for new services; document changes in version tab QLibrary-v2025-Q4.

Hold libraries stable within each quarter. Changing prompts mid-cycle invalidates trend comparison — see share of AI voice measurement.

1.2 Six-platform resample (2–3 hours DIY)

For each prompt × platform:

Platform	Settings to document
ChatGPT	Model version, browsing on/off
Gemini	Standard consumer path
Claude	Default web access state
Perplexity	Default mode
Grok	If relevant to audience
Google AI Overviews	Signed-out or typical user

Log:

• Named? Y/N
• Position if listed (1st, 2nd, 3rd)
• Themes quoted ("financing," "family-owned," "24/7")
• Citations (URLs) if shown
• Wrong facts about you or competitors

Compute:

• Mention rate = prompts naming you ÷ total prompts, per platform
• Platform coverage = count of platforms with ≥1 mention
• SOAV vs 3–5 tracked competitors — benchmarking guide

1.3 Trend comparison (15 min)

Compare to prior quarter:

Metric	Q-1	Q0 (this audit)	Δ
ChatGPT mention %
Gemini mention %
Perplexity mention %
Combined SOAV
Platform coverage /6

Flag Δ ≥ 10 points on any platform for root-cause review in Workstreams 2–5.

AIrecommend.ai scan/rescan automates this matrix — useful when manual sampling drifted or multi-location volume exceeds staff capacity.

Workstream 2 — Accuracy and wrong-fact repair

AI ** omits** unresolved entities and hallucinates from stale fragments. Accuracy pass before content spend.

2.1 NAP crosswalk (45 min per location)

Build a row for every authoritative source:

Source	Name	Address	Phone	URL	Hours	Status
GBP
Apple BC
Bing Places
Yelp
Website footer
Schema JSON-LD
Top industry directory

Pass criteria: Exact match on name, address, phone per NAP consistency guide. Service area language consistent.

Fail actions: Merge duplicates, claim unclaimed profiles, update schema, submit Apple BC corrections.

2.2 AI wrong-fact log review (30 min)

From Workstream 1 sampling, list every incorrect AI statement about you:

• Wrong phone / address / hours
• Services you do not offer
• Conflation with competitor name
• Closed-permanently when open

For each, hypothesize source of truth conflict — outdated directory, old press, employee-created GBP.

Prioritize fixes that appear on multiple platforms — likely core graph issue.

Deep process: AI reputation repair.

2.3 Closure / move / rebrand check (15 min)

Any location changes this quarter?

• GBP moved pin verified
• 301 redirects for old location pages
• Schema address updated
• Apple BC location status correct
• Review generation uses new name consistently

Rebrands without graph updates are a top cause of zero mention rate despite strong reviews.

Workstream 3 — Entity and technical health

3.1 Schema validation (30 min)

• LocalBusiness / ProfessionalService JSON-LD on homepage or location page
• @id stable URL
• sameAs includes GBP, Apple, major social
• areaServed matches real geography
• Service catalog reflects current offerings — not 2022 menu

Validate with Google Rich Results Test and schema linter. Errors fixed before new markup added — structured data guide.

3.2 llms.txt and robots (15 min)

• /llms.txt exists, factual, points to canonical service URLs
• No accidental Disallow: / for relevant AI crawlers where policy allows indexing
• Key service pages return 200, not redirect chains

Checklist: llms.txt, schema, robots.

3.3 Directory graph completeness (45 min)

Quarterly re-verify claims on:

• Google Business Profile (primary)
• Apple Business Connect — often missed, high Siri/Apple Intelligence leverage
• Bing Places
• Category-critical directories (Avvo, Healthgrades, Angi, etc.)
• BBB if category norms expect it

Log unclaimed and duplicate counts. Target zero duplicates next quarter.

Entity depth: building entity authority.

Workstream 4 — Reviews and theme alignment

Reviews remain the strongest universal signal for local AI — 2026 review guide.

4.1 Velocity and rating snapshot (15 min)

Metric	Prior quarter	This quarter	Δ
Google review count
Average rating
Reviews last 90 days
Response rate %

Benchmark: Ethical steady velocity beats burst campaigns. No gating, no incentives violating platform policy — right way to ask.

4.2 Theme extraction (30 min)

From last 20–50 reviews, tag recurring nouns/adjectives:

• Service themes: "same-day," "explained options," "cleaned up"
• Staff names if repeatedly praised
• Differentiators competitors lack

Compare to AI theme logs from Workstream 1. Misalignment? AI may be reading competitors' reviews instead — increase velocity and coach customers to mention real experience details (without scripting fake reviews).

4.3 Response audit (20 min)

• Negative reviews responded within 7 days
• Responses factual, non-defensive, no HIPAA/legal violations
• Owner responses include correct business name spelling

AI systems read owner responses as fresh corroboration of operational tone.

Workstream 5 — Competitive intelligence

5.1 Competitor SOAV table (30 min)

Using same prompt library:

Business	ChatGPT	Gemini	Perplexity	Combined SOAV
You
Competitor A
Competitor B
Competitor C

Note platform specialists — who wins only on Perplexity vs only on ChatGPT. Low cross-platform overlap is normal — 11% problem.

5.2 Citation gap analysis (30 min)

When Perplexity or browsing ChatGPT cites URLs, log domains:

• Competitor domains cited you lack
• Directories cited instead of your site
• Local media or study pages

Action mapping:

Gap type	Typical fix
Directory dominance	Claim, reviews, completeness
Competitor blog/FAQ	FAQ schema + buyer-intent content
Data study cited	Dominance-tier local research
Review count delta	Ethical velocity program

5.3 New entrant scan (15 min)

Search prompts for businesses you have not tracked before. Add emerging rivals to next quarter's SOAV set.

Workstream 6 — Attribution and next-quarter plan

6.1 First-party attribution (30 min)

Train front desk / CRM:

• "How did you hear about us?" — include ChatGPT, Gemini, Perplexity, AI Overview as selectable options
• Log 90-day count of AI-attributed leads
• Compare to mention rate trends — divergence may indicate zero-click wins without site visits — zero-click guide

6.2 Prioritized action queue (30 min)

Score fixes Impact × Effort:

Priority	Action	Owner	Due
P0	Merge duplicate GBP
P1	Apple BC claim
P2	FAQ schema on top 3 services
P3	Review request SOP refresh

Rule: No P3 content until P0–P1 accuracy passes.

Cap queue at 5–7 items per quarter per location — finish beats infinite backlog.

6.3 Budget alignment (15 min)

Map queue to resources:

Band	Typical quarterly spend
DIY time	Owner + coordinator hours
Tools	Scan, schema, review platform
Agency	AIrecommend.ai Growth $4,997/mo or Dominance $9,999/mo for multi-location execution

Reference: local AI marketing budget 2026.

Printable master checklist

Copy into Notion, Sheets, or print for field teams.

Measurement

• Prompt library versioned and stable
• Six-platform sample complete
• Mention rates calculated per platform
• SOAV vs competitors calculated
• Quarter-over-quarter delta flagged

Accuracy

• NAP crosswalk — all sources match
• Duplicates identified for merge
• AI wrong-fact log with source hypotheses
• Move/rebrand redirects verified

Entity

• JSON-LD validates without errors
• llms.txt current
• Apple BC + Bing claimed
• Category directories claimed

Reviews

• 90-day velocity logged
• Theme tags aligned with AI echo themes
• Response SLA met

Competition

• SOAV table updated
• Citation gaps mapped to fixes
• New entrants added to tracker

Planning

• AI-attributed calls logged
• P0–P3 queue assigned with owners
• Next audit date calendar-held

Multi-location adjustments

For 2–10 locations:

• Run full audit per location — mention rates vary wildly by geo
• Roll up SOAV by market, not brand average alone
• Centralize schema templates; localize areaServed, phone, address
• Apple BC often missed per-site — audit each

For 10+ locations:

• DIY quarterly audits collapse in spreadsheets — automate sampling
• Tier locations: top revenue markets monthly, long tail quarterly
• Dominance-tier wrong-fact tracing at scale

Integrating with SEO and GBP quarterly reviews

Merge calendars — many checks overlap:

SEO quarterly	AI quarterly add-on
Rank tracking	Mention rate tracking
Content refresh	FAQ aligned to AI themes
Backlink audit	Citation domain audit
GBP insights	AI wrong-fact + Apple BC
Technical crawl	llms.txt + schema for AI

Framework: AEO vs GEO vs SEO.

When quarterly is not enough

Move to monthly measurement if:

• Competitor SOAV shifted ≥15 points last quarter
• Active remodel / rebrand / new location
• Dominance-tier content or press launches
• Category known for fast model retrieval changes (legal, medical, emergency trades)

Measurement can be monthly; full six-workstream audit still suffices quarterly unless accuracy failures are chronic.

Common audit mistakes

Mistake: Prompt library churn. Different questions each quarter — trends meaningless.

Mistake: Single-platform obsession. ChatGPT-only blind to Perplexity gains — platform comparison.

Mistake: Skipping Apple BC. Repeated in every audit because operators forget — high leverage.

Mistake: Audit without approval queue. Listing fixes stall in "marketing will do it."

Mistake: Chasing guarantees. Vendor promises of placement invalidate the audit's purpose.

Mistake: No competitor set. SOAV requires tracked rivals — update list quarterly.

Case pattern — audit catches silent decay

Composite example:

Q2: Mention rate 28% combined, SOAV leader.

Q3 audit skipped.

Q4 audit: Mention rate 14%. Root causes found:

1. Former manager's GBP still claimed with old phone
2. Competitor published neighborhood service pages with FAQ schema
3. Review velocity dropped during staffing shortage
4. Perplexity citing competitor's local study

Q1 plan: P0 GBP reclaim, P1 review SOP, P2 FAQ deployment, P3 study scoping for Dominance tier.

Realistic expectation: Mention recovery over 2–3 quarters — not instant.

Bottom line

A quarterly AI visibility audit is six workstreams — measure mentions and SOAV, fix accuracy, validate entity health, align reviews, benchmark competitors, plan attribution-backed actions. It is the difference between knowing AI is your channel and guessing why leads dipped.

Use this checklist every quarter. Automate sampling where human drift hurts consistency. Fix NAP before blogs. Rescan trends, not anecdotes.

No audit guarantees ChatGPT will name you next month. Every skipped audit guarantees you will not know when it stops.

Free scan baseline · AI visibility tracking · First 90-day AEO roadmap.

---

Frequently asked questions

How often should local businesses audit AI visibility?

Quarterly at minimum for stable markets; monthly for competitive categories or active AEO programs. Model updates and competitor review velocity can shift mention rates faster than traditional SEO rankings.

What platforms should a quarterly audit cover?

At least six — ChatGPT, Gemini, Claude, Perplexity, Grok, and Google AI Overviews — using the same buyer-intent prompt library each cycle for comparable trends.

Can I run a quarterly AI audit myself?

Yes. You need a fixed prompt library, a logging spreadsheet, listing access, and 3–5 hours per location per quarter. Automation via AIrecommend.ai rescans reduces manual drift and improves consistency.

What is the first thing to fix when an audit finds problems?

Factual accuracy — wrong NAP, duplicate listings, and incorrect hours — before content or link-building. AI omits or misstates businesses it cannot resolve confidently.

Does AIrecommend.ai replace the quarterly audit?

It automates sampling, competitor comparison, and gap mapping to Growth Engine modules — but you still approve listing changes, review responses, and strategic priorities. The audit is a process, not a one-time report.