How to Tell If ChatGPT and Claude Are Recommending Your Business (And What to Do If They're Not)

Your next customer is not typing your category into Google. Not exclusively. They are asking ChatGPT, Claude, or Perplexity a question that sounds a lot like "who's the best [your category] in [their city]?" or "which [your product type] should I buy for [their use case]?" The AI answers. The customer picks from the AI's shortlist. You are either on that shortlist or you are not.

Most businesses have no idea whether they are on it. They have not checked. They are not tracking it. The category most affected by AI search — local service businesses, mid-market B2B, professional services, retail brands — is also the category with the least AI-visibility instrumentation. You are being ranked or excluded by a system you have never opened.

This post is a 15-minute diagnostic anyone can run today, plus the five fixes that move the needle if your business is invisible. No paid tools required. If you already know your business is invisible in AI search, skip to the fixes. If you don't know — and 90% of business owners I talk to have no idea — start with the diagnostic.

How do I check if ChatGPT is recommending my business right now?

Open ChatGPT in an incognito or logged-out browser session, then ask the same question your buyers ask. Do it three ways: the generic category query, the category-plus-location query, and the buyer-intent query. Write down whether your business is named, listed among competitors, or absent entirely. That's your baseline. Fifteen minutes end to end.

The reason for incognito matters: if you use your own ChatGPT account, your prior searches and preferences bias the response toward things you have already talked about. The buyer who has never heard of you is starting from zero. You need to see what they see.

Run each query in ChatGPT, Claude, and Perplexity. If your business shows up in one but not the others, the fix is model-specific and I'll cover it below.

What prompts should I actually try?

Use three prompt shapes: the generic recommendation, the specific-need query, and the shortlist request. Each one probes a different way the AI decides who to name, and each one maps to a different stage of your buyer's real decision.

Here are the three prompt shapes with concrete examples. Substitute your category and city.

• Generic recommendation. "Who's the best HVAC company in Denver?" or "What's a good B2B SEO consultant for a healthcare company?"
• Specific-need query. "I need an HVAC company in Denver that services Trane heat pumps and can come out this week — who should I call?" or "Which SEO consultant should I hire if my site is losing citations in AI Overviews?"
• Shortlist request. "Give me a shortlist of five HVAC companies in Denver, with a one-sentence description of each."

The shortlist request is the highest-signal one. If you don't appear in a shortlist of five, you are functionally invisible to that model. Being #6 in Google's organic list still gets you some traffic. Being #6 on ChatGPT's shortlist gets you zero.

Run all three prompts across all three models (ChatGPT, Claude, Perplexity). That's nine tests. Ten minutes.

What does "being recommended" actually look like?

There are three outcomes worth distinguishing: named as a top recommendation, mentioned in a longer list, or absent entirely. Only the first two count as visibility, and only the first counts as a strong recommendation. If your business is mentioned in a paragraph that ends with "and other local providers," that is still absence.

Here is the pattern I look for when I audit a client's AI visibility:

Outcome	What it looks like	What it means
Named first, with description	"For Denver HVAC, Cool Comfort Heating is a strong pick — they specialize in…"	Strong recommendation. AI has confidence in your entity.
Named in top 3–5 shortlist	"Consider: [Comp A], Your Business, [Comp B]…"	Solid visibility. You're eligible for the pick.
Named in a longer paragraph	"…other providers include Your Business, [Comp X], [Comp Y]…"	Weak visibility. You appear but rarely close.
Not mentioned at all	AI names 5 competitors, none of them you	Invisible. This is where most businesses are.

If you are in the bottom two rows across all three models, your AI-recommendation surface is broken. That is what this post is about fixing.

Why isn't my business showing up?

Five reasons, in order of frequency. If your business is invisible, at least three of these apply. The good news is all five are fixable, and none of them require running paid ads or building backlinks the traditional way.

One: your business has no entity clarity. AI models identify businesses as entities the way a librarian identifies a book — a title, an author, a topic, a place. If your website doesn't declare structured data (JSON-LD schema) that says "we are a business named X, in category Y, located at Z, offering services A/B/C," the AI has to guess. Most guesses are wrong.

Two: your content answers no specific questions. LLMs pull from sources that contain direct, extractable answers to buyer questions. If your site is a series of marketing paragraphs about "excellence" and "trusted partnership," there is nothing for a model to quote. If your site has a page titled "How much does HVAC repair cost in Denver?" with a direct answer in the first paragraph, that page can get cited.

Three: your name doesn't appear alongside category terms in enough contexts. AI models learn associations from co-occurrence in training and citation data. If your business name only appears on your own website and a handful of directory listings, the model has thin ground to associate you with the category. Businesses that get recommended have their name appearing in industry publications, review sites, local press, forum threads, and third-party guides.

Four: your reviews are thin or invisible. Real review volume on Google Business Profile, Yelp, and category-specific platforms (Angi, HomeAdvisor, Clutch, G2, whatever fits your category) is a heavy signal. AI models treat 200 recent reviews as a signal your business is real, active, and reasonably competent. Twenty reviews from 2019 gets you nothing.

Five: you have no Wikipedia-style third-party summary of your business. For mid-market and above, having a Wikipedia entry is the single strongest AI-visibility signal that exists. For small businesses, you can't get one. But you can get the next-best-thing: a well-structured "about" page with Organization schema, plus mentions in industry-authority sources that summarize what you do.

How do you fix each one?

Fix them in the order they appear above, because each one compounds the next. Entity clarity is the foundation — without it, none of the other fixes register at full strength. Here is the concrete work for each.

Fix entity clarity: add Organization + LocalBusiness (or ProfessionalService) JSON-LD schema to your site. Every page. Declare business name, address, phone, service area, category, hours, price range, sameAs links to your social profiles and directory listings. This is a one-hour engineering task that most small business websites never get. It is table stakes for AI visibility in 2026.

Fix the content answer problem: write question-headed pages. Every buyer question in your category becomes an H2 on a page, followed by a 40–60 word direct answer, followed by supporting detail. Not "Our HVAC Services" as an H2. "How much does an HVAC repair cost in Denver in 2026?" as an H2. The tool I built at /products/seo-geo-aeo-checker will score whether any given page on your site has this structure — start there and fix the pages that score under 70.

Fix the co-occurrence problem: get named in third-party sources. This means guest posts on category publications, being quoted in journalist queries (HARO, Qwoted, Featured), being interviewed on niche podcasts, sponsoring or attending local industry events, contributing to open-source projects if you're in B2B tech, publishing case studies that get shared. Each individual placement is a small signal. Ten of them over a quarter is a real signal. Twenty over a year is category-changing.

Fix the review problem: run a real review-generation program. Ask every closed customer for a review within 48 hours of the transaction. Use SMS or email. Make it one click. Not "would you leave us a review?" but "here's the link — takes 30 seconds — thanks." A business that gets 5 new reviews per month for a year has 60 more reviews than a business that runs no program. That gap is visible to AI models.

Fix the third-party summary problem: publish structured "about" content and get it cited. A well-written 800-word "about" page that clearly states who you are, what you do, who your customers are, how long you've been operating, and what makes your work different — with author markup, publication date, and structured data — is a real GEO asset. Then link to it from every third-party mention you can get. The goal is that when an AI model tries to identify you as an entity, it finds a coherent summary in your voice AND coherent summaries in other people's voices, all pointing to the same facts.

Do the same checks work for Claude and Perplexity?

Yes, but each model has quirks. ChatGPT tends to over-index on brand recall and generic authority. Claude tends to over-index on structured content and clear source signals. Perplexity tends to over-index on recency and citation link quality. Same underlying game, different weightings — which means a business that scores well in one may miss in another.

Practical differences worth knowing:

• ChatGPT will name well-known brands even when the query is local. If you compete against a national franchise with strong brand recognition, expect ChatGPT to favor them regardless of location.
• Claude rewards clear author markup, publication dates, and Q&A-formatted content more than the others. If your site has good FAQPage schema and question-driven H2s, you'll typically appear stronger in Claude than in ChatGPT.
• Perplexity shows its citation sources inline, which means you can see exactly which pages it pulled from. Use this as a direct signal — if it cites a competitor's blog post, that post is worth studying and beating.
• Google AI Overviews is a fourth surface worth checking. It sits at the top of Google search results now, pulls from a small citation set (usually 3–5 sources), and follows different rules than the chat models. Being cited in AI Overviews is often more valuable than ranking #1 organically, because AI Overviews appear above the ten blue links.

Run all four surfaces. Don't optimize only for whichever one you happen to use personally.

How often should I re-check?

Monthly at minimum, weekly during any period of active fixing. AI models retrain and rebalance, index and re-index sources, and change their citation behavior. A business that was cited last month may not be this month. A fix you made last week may not show up in the model's answer for another two to four weeks. Cadence matters.

Practical routine: on the first Monday of each month, run the nine-test diagnostic (three prompts × three models). Log results in a spreadsheet. Track your appearance rate (out of 9 tests, how many named you?) over time. If the number stays flat or drops after you've done the five fixes, something upstream is wrong and it's audit time.

The bigger picture

The mistake most small and mid-market businesses are making right now is treating AI search as a novelty. It is not. It is where a growing share of category-defining decisions is getting made, and the businesses that get ahead of it in 2026 are going to compound that advantage for years, because AI models weight authoritative, structured, well-cited sources — and the businesses that build those signals first accumulate the advantage.

The businesses that don't will notice, sometime in late 2026 or 2027, that their inbound has slowly gone quiet. They'll blame the economy, blame their agency, blame Google. The real answer will be that their buyers stopped asking Google and started asking ChatGPT, and their business was never in the answer.

Run the diagnostic. Fix the misses. Re-check in a month.

If you want the automated version of this diagnostic — the tool that scores whether any given page on your site is structured to be cited by AI engines — it is free at /products/seo-geo-aeo-checker. If you want the deeper story on why SEO, GEO, and AEO are three different jobs, that is in the companion article. And if you want me to run a full AI visibility audit on your business — the manual version, with actual queries tested across all four surfaces and a prioritized fix list — book a consult and we'll scope it.