Why ChatGPT Does Not Recommend Your Business — A Diagnostic Guide

The moment you realize you are invisible

A prospect says, "I asked ChatGPT who to use — they never mentioned you."

You open ChatGPT, type your business name, get a polite summary. Relief — until you try what the prospect actually asked: "Who's the best [your category] in [your city]?"

Three competitors appear. Your name does not.

That gap is more common than most local owners expect — and more fixable than panic suggests, within limits no vendor controls. ChatGPT is a third-party product. OpenAI does not publish local ranking factors. Nobody guarantees organic recommendations.

What you can do is diagnose signal gaps, fix verifiable public evidence, measure mention rates across engines, and stop confusing SEO success with AI visibility.

This article is a structured diagnostic for local and service businesses invisible on buyer-intent prompts — starting with ChatGPT because it carries the most conversational mindshare, then widening to the full platform panel AEO and GEO programs require.

Confirm the problem is real — not a bad query

Before rebuilding your marketing stack, validate the symptom.

Run buyer-intent prompts, not brand queries:

• "Recommend a [category] in [city] for [specific need]"
• "Who should I call for [urgent service] near [neighborhood]?"
• "Best [specialty] with good reviews in [metro]"

Use fresh chats to reduce session bias. Note whether browsing is enabled — answers can differ when ChatGPT retrieves live web and directory data versus relying on training memory.

Log for each prompt:

1. Which businesses were named (up to three)
2. Stated reasons — reviews, years in business, services
3. Any factual errors about you or competitors
4. Whether the model refused to recommend anyone

Repeat on Gemini, Perplexity, and Claude minimum. ChatGPT-only diagnosis misses platform blind spots — cross-engine overlap is low in many samples (~11% shared domains in some industry analyses). Read: The 11% problem.

Structured baseline: How to check what ChatGPT says · Free six-platform scan.

If you appear on brand queries but not category queries, you have a recommendation-layer gap, not total invisibility. The fixes below still apply — emphasis differs.

Diagnostic framework — eight failure modes

Work through these in order. Earlier modes block later ones — schema on a business with twelve reviews rarely breaks through a competitor with four hundred.

Failure mode 1 — Insufficient review density

ChatGPT and peer engines often cite review counts and themes when explaining recommendations. Thresholds are not published, but sub-critical volume produces silence or generic "search locally" advice.

Symptoms:

• Named competitors have 3–10× your Google review count
• Model cites "highly reviewed" or "many positive reviews" for others, not you
• You have high stars but low count (e.g., 5.0 with nine reviews)

Diagnosis steps:

Compare your Google review count and recency to the median of named competitors on the same prompts.

Scan review text for theme coverage — emergency, financing, communication, specialty services — not only stars.

Fixes (no guarantees):

Ethical review velocity through completed jobs — not gating, not incentives violating Google policy. Guide: Google reviews the right way.

Respond to reviews; freshness signals active operations.

Target theme gaps — if prompts stress "same-day," reviews should organically mention speed from real customers.

Timeline: Review accumulation is months, not days. Measure monthly mention rate, not daily ChatGPT checks.

Failure mode 2 — NAP inconsistency and listing drift

Entity resolution systems merge or split businesses based on name, address, phone agreement across sources. Drift makes you a ambiguous record — easy to skip when models pick clear entities.

Symptoms:

• Old phone on Yelp, new phone on site
• Suite numbers differ across directories
• DBA vs legal name chaos
• Unclaimed duplicates on Bing or Apple

Diagnosis steps:

Export NAP from GBP, website footer, top five directories, and Apple Business Connect.

Highlight mismatches in a diff table.

Search your phone number — do unexpected listings appear?

Fixes:

Claim and correct Apple Business Connect — common blind spot: Apple Business Connect guide.

Standardize NAP; update directories in priority order: Google, Apple, Bing, Yelp, vertical dirs (Healthgrades, Avvo, Angi).

Remove or merge duplicate listings through platform workflows.

Timeline: Weeks for directory propagation; resample AI after major NAP cleanup.

Failure mode 3 — Google Business Profile incompleteness

Even ChatGPT, which is not Google, often ingests public web and directory corpora where GBP-derived or Google-indexed content appears. Gemini and AI Overviews lean heavily on Google's local graph.

Symptoms:

• Missing services list or generic categories only
• No Q&A answered; stale photos
• Service-area business with unclear coverage
• Hours wrong or holiday closures missing

Diagnosis steps:

Compare GBP completeness score and fields to named competitors.

Read GBP services against your actual job mix — are specialty services listed?

Check if competitors have regular posts and you do not.

Fixes:

Complete primary and secondary categories, services, attributes, service area.

Answer Q&A with buyer-language questions.

Post updates on seasonality, hiring, community — operational signals.

Align GBP with site schema and llms.txt.

Failure mode 4 — Weak or absent entity profile on your site

Models browsing your domain need crawlable, factual HTML plus structured identity. Thin sites with marketing fluff and no location scope give little to ground.

Symptoms:

• Homepage is brochure copy without address, phone, service area
• No location or service pages
• Missing JSON-LD LocalBusiness; no llms.txt
• robots.txt blocks key pages

Diagnosis steps:

View source — is JSON-LD present and accurate?

Fetch /llms.txt — does it exist?

Run Rich Results Test and manual crawl of service URLs.

Fixes:

Deploy LocalBusiness subtype schema, llms.txt, and service pages with geography — technical guide: Structured data for AI assistants · llms.txt checklist.

Ensure robots.txt allows public pages.

Timeline: Days to deploy; AI pickup varies by retrieval path.

Failure mode 5 — Query–theme misalignment

You may be strong on generic prompts but invisible when buyers add modifiers — emergency, pediatric, commercial, cosmetic, affordable.

Symptoms:

• Named on "best plumber [city]" but not "same-day water heater [city]"
• Reviews rarely mention the adjectives buyers use in prompts
• GBP services omit specialties you actually perform

Diagnosis steps:

Segment your prompt library by intent cluster; calculate mention rate per cluster.

Compare review n-grams to prompt modifiers.

Fixes:

Add honest FAQ content and schema for specialty services.

Build review themes organically through delivered work matching those specialties.

Expand GBP services and service page copy — factual, not stuffed.

Failure mode 6 — Competitor signal dominance

Sometimes you are not "broken" — competitors are ** exceptionally dense** on public evidence: 900 reviews, press coverage, data studies, multi-directory presence.

Symptoms:

• ChatGPT names the same two market leaders on most prompts
• You appear on long-tail niche prompts only
• Perplexity cites competitor URLs repeatedly

Diagnosis steps:

Build a signal scorecard — reviews, directories claimed, schema yes/no, recent press, citable pages — for you vs top three named competitors.

Identify the largest gap class, not the easiest task.

Fixes:

Long-game review and listing parity where realistic.

Citable data studies and merit-based press for Perplexity-heavy journeys.

Niche dominance — own specific intent clusters rather than vague "best in city" head terms.

Accept that SOAV may stay below market leaders while mention rate rises from 0% to 25% — still revenue meaningful.

Measure share of AI voice: measurement guide.

Failure mode 7 — Wrong or hallucinated facts about you

Paradoxically, some businesses are ** omitted because models "know" wrong things** — closed, wrong category, outdated phone — and confidence filters skip unreliable entities.

Symptoms:

• ChatGPT states incorrect hours or services for your business when asked directly
• You appear inconsistently — named once, omitted next session
• Third-party directories show obsolete data

Fixes:

Trace sources; fix upstream listings and schema: AI reputation repair.

Queue corrections through approval workflow — do not spam model feedback forms expecting instant fixes.

Timeline: Accuracy repairs can improve direct queries before category recommendations catch up.

Failure mode 8 — Platform-specific blind spots

You may be visible on Gemini and invisible on ChatGPT — or the reverse. Single-platform obsession misallocates budget.

Symptoms:

• Six-platform scan shows 0% on two engines, 40% on another
• Sales hears "ChatGPT" but not "Gemini" — your gap may still be wide on Google-adjacent surfaces

Fixes:

Platform-specific remediation — Google ecosystem for Gemini/AI Overviews; citable pages for Perplexity; entity + reviews universal baseline for ChatGPT.

LLM SEO is the umbrella discipline — measure all major engines.

Decision tree — where to start this week

START: Invisible on buyer-intent prompts?
│
├─ Review count < 50% of competitor median?
│   └─ YES → Review Engine priority (ethical velocity)
│
├─ NAP mismatches on ≥2 major directories?
│   └─ YES → Listings audit + Apple BC claim
│
├─ GBP <80% complete vs competitors?
│   └─ YES → GBP services, Q&A, posts, hours
│
├─ No JSON-LD / llms.txt / service pages?
│   └─ YES → Entity Profile deployment
│
├─ Wrong facts when asked directly?
│   └─ YES → Accuracy repair workflow
│
└─ Visible on some platforms only?
    └─ YES → Platform-specific fixes + full panel resample

One primary lever per 30-day sprint avoids scattered half-fixes.

What not to do

Buy "ChatGPT placement packages." No ethical vendor controls OpenAI outputs.

Keyword-stuff GBP or schema. Suspensions and entity noise hurt long-term.

Review gating. Policy violations; unnatural distributions.

Obsess over single screenshots. Noise dominates short samples.

Abandon SEO entirely. Strong pages still feed retrieval and convert clickers.

Assume training-data immortality. New businesses win through browsing and directories when memory lags.

Setting honest expectations

Fixes improve probability of mentions by strengthening evidence models already use. They do not contractually bind ChatGPT.

Model updates can reset trends.

Competitors respond to the same playbook.

Some categories in some markets have entrenched leaders — mention rate gains, not #1 SOAV, may be the realistic win.

Report monthly mention rate and share of AI voice with platform breakdowns. Tie call tracking to operational outcomes.

Case patterns (composite scenarios)

Pattern A — Strong SEO, zero ChatGPT. 200 reviews, page-one rank, thin schema, Apple unclaimed. Fix entity + Apple; mention rate 0% → 22% over four months on sampled prompts — illustrative, not guaranteed.

Pattern B — New business, total silence. 14 reviews, perfect site, full schema. Review density and directory breadth insufficient. Focus reviews + vertical directories; schema already done.

Pattern C — Named with wrong phone. Accuracy repair + NAP sync; direct queries correct first; category mentions follow slowly.

Pattern D — ChatGPT yes, Perplexity no. Add citable FAQ and data content; Perplexity moves while ChatGPT stable.

Use patterns as hypotheses — your scan data decides.

Monthly operating rhythm

Week	Activity
1	Resample prompt library; update mention rate + SOAV
2	Execute one signal sprint (reviews, listings, or entity)
3	Accuracy check — direct brand + category prompts
4	Leadership readout — trends, not guarantees

Automate sampling where possible: AI visibility tracking.

When to escalate to professional help

DIY diagnosis works for single-location SMBs with time. Consider AEO / GEO programs when:

• Multi-location NAP complexity
• Accuracy crises with revenue impact
• Competitive SOAV tracking across 20+ prompts
• Need approval-gated fix execution (Entity Profile, press, studies)

Reject providers promising guaranteed ChatGPT recommendations.

Summary

ChatGPT does not recommend your business when public evidence — reviews, listings, entity clarity, theme alignment — falls below competitive thresholds or conflicts across sources. The diagnostic path is measurable:

1. Confirm with buyer-intent prompts across multiple platforms
2. Score eight failure modes against named competitors
3. Fix the highest-leverage verifiable inputs first
4. Resample monthly without placement guarantees

Start: Free scan · How AI assistants choose businesses · AEO vs GEO vs SEO.

Questions to ask your marketing team today

Before the next budget meeting, ask:

1. What is our mention rate on ten buyer-intent prompts — any number, even zero?
2. Which platforms never name us?
3. What is the median Google review count of businesses ChatGPT names instead of us?
4. Does our NAP match on Apple Business Connect and Yelp?
5. When ChatGPT describes us directly, are facts accurate?
6. Do we have JSON-LD and llms.txt deployed from verified intake, not scraped guesses?
7. Are we reporting Overview mention rate on Google, not only map pack?

If most answers are "we don't know," diagnosis has not started — only anxiety.

Integration with existing SEO retainers

Many businesses pay for local SEO while AI invisibility persists. Merge conversations with your SEO partner:

• Request AI mention sampling alongside rank reports
• Map citation work to entity resolution, not only link equity
• Align content calendar with buyer-intent FAQ gaps surfaced in scans
• Share AEO vs GEO vs SEO framing so both sides use the same vocabulary

SEO partners who refuse to discuss mention rates are not wrong about SEO — they may simply be out of scope. Add scope or add a specialized AEO measurement layer.

Visibility on the recommendation layer is earned through signal density and consistency — not purchased placement in a black box.