---
title: "ChatGPT Cited Your Website. It Still Recommended Someone Else."
description: "A citation proves your page helped answer a question—not that your business made the shortlist. Here’s what to fix and what an AI-search provider should report."
url: "https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/44d22376-dbef-4159-9018-ec7636672b48.png"
published: "2026-11-20T09:27:58.569Z"
modified: "2026-11-20T09:27:58.651Z"
---

November 20, 2026 · 10 min read

# ChatGPT Cited Your Website. It Still Recommended Someone Else.

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Cartoon of three rival cleaning crews on a winners' podium while a lone cleaner sits below holding a tiny 'Cited' ribbon: credited as a source, but not chosen.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/44d22376-dbef-4159-9018-ec7636672b48.png)

In this article

1. [Three kinds of visibility—and only one is a recommendation](#three-kinds-of-visibilityand-only-one-is-a-recommendation)
2. [The buyer asked “who,” not “how”](#the-buyer-asked-who-not-how)
3. [Three checks between useful information and buyer fit](#three-checks-between-useful-information-and-buyer-fit)
4. [What an AI-search provider should actually fix](#what-an-ai-search-provider-should-actually-fix)
5. [Keep a shortlist log, not just a citation chart](#keep-a-shortlist-log-not-just-a-citation-chart)
6. [Apply the model to a different business](#apply-the-model-to-a-different-business)

**ChatGPT can cite your website and still recommend three competitors.** Your page helped explain the work; their businesses made the shortlist.

Say you run a commercial cleaning company in Austin. Someone asks ChatGPT for the best cleaner for a medical office. The answer borrows your guide’s cleaning checklist, links to your page, then names other firms as the options worth calling. A citation report records a win. I would want to read the rest of the answer.

Being useful as a source and being selected as a supplier are different outcomes. An AI-search service that counts the first without checking the second can mistake publishing success for commercial visibility. To see where the work actually sits, I’ll follow that Austin cleaner from the footnote to the shortlist.

## Three kinds of visibility—and only one is a recommendation

ChatGPT can do three different things with your business. I keep them separate:

- **Citation:** the answer [uses your page as a source](https://onely.com/blog/why-ai-chatbots-cite-you-but-dont-recommend-your-brand/) and attributes it with a link.
- **Mention:** your business name appears in the answer text.
- **Recommendation:** your business is presented as an option that fits the buyer’s request.

These can overlap. They do not have to. A mention alone is not an endorsement, and a source link does not mean the answer recommends the company behind it.

In the Austin example, your guide explains how medical offices should be cleaned. ChatGPT cites the checklist, then recommends other firms whose information shows medical clients, compliance steps, and outside reviews tied to that job.

**You won the footnote. They won a place on the shortlist.** Neither result, by itself, tells us who booked the job. But only one tells us that the answer presented your business as a supplier to consider.

That is the distinction I want a report to preserve. Put citations, mentions, and recommendations in separate columns before discussing whether visibility improved.

## The buyer asked “who,” not “how”

Your cleaning guide answers _how_ medical cleaning should work. The buyer asked _who_ should do it for their office.

The checklist helps write the explanation. The shortlist needs evidence about the supplier:

- Which facilities does the company serve?
- What compliance steps does it follow?
- What hours does it cover?
- Who else confirms that it does this work well?

A good explanation does not necessarily answer any of those questions. That is why another checklist article may earn another citation without making the business a stronger candidate.

The split shows up in the reporting, too. One analysis found that [about 62% of AI citations linked to a domain without naming the brand](https://citemetrix.com/blog/ghost-citations-62-percent-no-brand-mention/). Your URL can sit in the footnotes while your name never appears in the text. The reverse happens as well: [in one B2B test, 66% of recommendations came without a citation to the recommended brand’s own site](https://eclincher.com/blog/how-chatgpt-recommends-businesses-signals/).

Those are different tests, not a universal formula. They illustrate the distinction: an answer can use your information without choosing you, or choose you using information published elsewhere.

**The commercial task is not just to explain the service. It is to connect your business to the buyer’s requirements.** That gives us a working model for deciding what to fix.

## Three checks between useful information and buyer fit

Start from the fact everything else follows from: ChatGPT cannot recommend a business as a fit without connecting that business to the buyer’s need.

I use three checks to examine that connection: access, identity and fit, and outside confirmation. This is a diagnostic model, not a claim that every answer follows a fixed internal sequence. A recommendation can draw on third-party sources without citing your website at all.

### 1. Can the search system reach the useful information?

For your own page to contribute as a source, it needs to be accessible. That sounds trivial until someone closes the gate during a site rebuild.

[OAI-SearchBot serves ChatGPT search, while GPTBot is used for training](https://aiseewhy.com/blog/gptbot-oai-searchbot-chatgpt-user-explained/). Their `robots.txt` settings are independent. Treating them as interchangeable can leave a site open to one use and blocked from another.

Say our Austin cleaner accidentally blocks the search crawler, or puts useful pricing information behind JavaScript the fetcher skips. The page cannot contribute the information it was supposed to supply if the search system cannot retrieve it.

That is an access problem. It is not yet an explanation for why another business made the shortlist, because outside sources may still describe both companies.

**Check retrieval before rewriting a page that may not be getting read.** Then check whether the page actually supplies the answer. Many citation fixes concern [a missing answer, not a missing keyword](https://groas.com/post/what-a-content-gap-actually-is-when-chat).

### 2. Does the information identify a supplier that fits?

Now suppose the cleaner’s guide is accessible and gets cited. It explains terminal cleaning for surgical suites, but never says whether the company serves Austin medical offices, which facilities it takes, what training its technicians complete, or what it will not do.

The method is clear. The commercial fit is not.

That gap belongs on a service page, not in another paragraph explaining why cleanliness matters. Our cleaner needs plain statements about its service area, facilities, hours, compliance steps, and limits. A buyer should not have to infer those facts from a general guide. Neither should the answer assembling the shortlist.

**Make the buyer, service area, and limits explicit.** Keep the useful guide, but give it a commercial counterpart that answers, “Is this company right for this office?”

This is work a citation-only report can miss. The guide is already doing its publishing job. The missing information concerns the business behind it.

### 3. Does anyone else confirm the fit?

A company’s own page can state what it does. Outside sources can help confirm that claim.

For the Austin cleaner, that layer might include medical-office testimonials, Google reviews, a listing with consistent categories and hours, or a local publication covering relevant vendors. These sources do a different job from a second blog post repeating the cleaning checklist.

Product-question research illustrates why looking only at brand pages misses part of the picture: [review and testing sites lead citations in one analysis](https://carsoncobb.com/blog/the-sources-chatgpt-cites-for-product-questions-2026/), while [AI shopping answers also draw heavily on YouTube, Reddit, and independent testers](https://cloro.dev/blog/ai-shopping-what-chatgpt-recommends-data-study/). Those are product studies, not a measurement of our hypothetical cleaner. The practical lesson is to inspect the outside sources in the actual answer rather than assume the company website supplies everything.

For local search specifically, a tracker of 200,000 non-branded prompts found that [business websites supplied 42% of citations, Google Business Profile supplied 28.6%, and Yelp about 9.5%](https://brightlocal.com/research/ai-local-search-data/). The confirming information sits partly off your site.

**Check whether outside evidence supports the same buyer fit your commercial page claims.** For our cleaner, that means medical-office work in Austin—not merely a collection of generic praise for being punctual.

## What an AI-search provider should actually fix

If I were hiring someone to help that cleaner get recommended, I would not start with their citation chart. I would ask them to show which of the three checks is failing, what they plan to change, and how they will test it.

That turns a vague visibility service into three inspectable jobs.

### Make the right pages accessible and useful

Can the engines fetch the pages that matter? Do those pages answer the buyer’s question in a retrievable passage?

The work includes checking `robots.txt` settings for OAI-SearchBot, reviewing server logs for what actually gets fetched, and publishing buyer-question pages that state service, area, fit, and limits outright.

I use this [method for testing ChatGPT versus Perplexity citations without grading your own homework](https://groas.com/post/how-to-build-a-chatgpt-perplexity-citation-test-that-doesn-t) because it makes the provider show its work across engines instead of screenshotting one friendly prompt.

**Ask for the retrieval problem and the proposed page change, not just a promise of more visibility.** If the guide is already cited, the provider should explain why editing it addresses the missing recommendation.

### Build evidence beyond the company website

The next job is to examine the reviews, listings, mentions, and category pages that support—or fail to support—the business’s claims.

For our cleaner, the same service area, categories, and hours should appear consistently wherever those facts are published. Relevant testimonials should make the medical-office fit clear. A generic listing and a specific commercial page should not describe what looks like two different businesses.

This is where I separate dashboards from execution. groas provides infrastructure for monitoring the market and running changes under your guardrails, with changes logged alongside their reasons. I would rather inspect that work than receive a monthly PDF reporting my footnotes back to me.

**The deliverable is a clearer, better-supported supplier profile—not another count of URLs.**

![A cited cleaning guide beside a shortlist naming competing suppliers](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/9b9a4335-a829-40b2-b2f1-b2b6c628a81b.png)

### Test questions that carry buying intent

Finally, test the questions a buyer would actually ask. For the Austin cleaner, that set could include:

- Best medical office cleaning in Austin.
- Who handles surgical suite turnover in Austin?
- What does compliant medical-office cleaning cost per month?

Run the same set across ChatGPT, Perplexity, and AI Overviews twice a month, and record what changes. Keep the questions consistent enough to compare the answers.

A citation tool can tell you whether your URL appeared. That alone cannot tell you whether your company became an option worth calling. I want to see whether the Austin business moves from source material to named candidate after the fix.

**Test the recommendation set, not just the presence of a link.** Otherwise, the provider is grading the wrong outcome.

## Keep a shortlist log, not just a citation chart

If you take one habit from this piece, take this log. For every commercially meaningful prompt, record six fields:

1. **Check date.**
2. **Exact prompt.**
3. **Engine and sign-in status.**
4. **Full answer, saved as text.**
5. **Pages cited.**
6. **Suppliers named, distinguishing recommendations from passing mentions.**

That small table separates visibility evidence from wishful thinking. It is also the discipline behind [measuring AI visibility without tracking fake rankings](https://groas.com/post/how-to-measure-your-brand-s-visibility-i).

Citations belong in one column. Named recommendations belong in another. Leads, booked jobs, and revenue stay in the CRM, separate until you can trace a buyer back to an AI answer. A shortlist appearance is commercially relevant visibility; it is not a sale.

For the cleaner, a useful report would show the original answer, the missing fit information, the changes made, and the later answers. Perhaps the guide remains cited while the company begins appearing among recommended suppliers. Perhaps nothing changes. Either way, there is something concrete to inspect rather than a celebratory arrow pointing at a footnote.

**When a provider sends a citation chart, ask for the shortlist column. If it is missing, the report shows source visibility—not evidence that buyers are being directed toward your business.**

![A notebook log recording prompts, answers, and named suppliers by date](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/6e1db13f-e306-4d26-9a4b-f1584833153c.png)

## Apply the model to a different business

Now take a new case: you sell scheduling software for dental practices. ChatGPT cites your help document on recall reminders but recommends three rivals.

The help document has done its job. I would not start by rewriting it.

First, check access to the commercial pages that should establish buyer fit. The help document’s citation tells us that document was usable; it does not settle whether the relevant product information is accessible.

Next, examine what those commercial pages actually say. Suppose the product is built for practices with two to ten chairs, serves Texas and Florida, and integrates with the two booking systems those offices use. Those facts belong in plain sentences, alongside what the product does not support. They should describe the product’s actual limits, not a wishlist assembled to match a prompt.

Then look for outside confirmation of that same fit: category listings, dental forums, and detailed practice reviews. Two or three relevant reviews describing the right kind of practice tell a different story from generic praise that the software is “easy to use.”

Test the buyer prompts again in two weeks. Read the shortlist column, not just the footnote column. If the product still does not appear, the citation remains a publishing result. It has not become a recommendation merely because the link survived.

**Being quoted means your information helped. Being recommended means the answer connected your business to that buyer.** That is why I would use groas’s market monitoring and logged execution to work on the second job, rather than treat a citation count as the finish line.

This will not be the right priority for everyone. If your buyers still call from Google Maps and your schedule is already full, spend the money on capacity. If buyers now paste a ChatGPT shortlist into the first call, ask for the shortlist log before you renew anything.

Footnotes do not book jobs. Names on a list do.

## Frequently Asked Questions

### Can ChatGPT cite my website and still recommend my competitors?

Yes. ChatGPT can use your page as a source and link to it, then name other businesses as the options worth calling. A citation proves your content helped answer the question, not that your business made the shortlist.

### What is the difference between a citation, a mention, and a recommendation in ChatGPT answers?

A citation means the answer used your page as a source and linked to it. A mention means your business name appears in the answer text. A recommendation means your business is presented as an option that fits the buyer's request. They can overlap, but a source link alone is not an endorsement.

### Why does ChatGPT use my content as a source but recommend other companies?

Because a guide answers how a service should work, but the buyer asked who should do it. To build a shortlist, the answer needs supplier evidence: which facilities the company serves, what compliance steps it follows, what hours it covers, and who else confirms it does this work well.

### How do I check whether ChatGPT can actually access my website?

Check robots.txt settings for OAI-SearchBot, which serves ChatGPT search, separately from GPTBot, which is used for training, because their settings are independent. Also review server logs to see what actually gets fetched, and make sure useful information is not hidden behind JavaScript the fetcher skips.

### What information should a service page contain so AI answers can recommend the business?

A service page should state plainly the service area, the facilities served, hours, compliance steps, and what the company will not do. A buyer should not have to infer these facts from a general guide. Keep the useful guide, but add a commercial counterpart that answers whether the company is right for that specific buyer.

### Does ChatGPT only use my own website when deciding which businesses to recommend?

No. Outside sources confirm what a company's own page claims. For local search, one tracker of 200,000 non-branded prompts found business websites supplied 42% of citations, Google Business Profile 28.6%, and Yelp about 9.5%. Reviews, listings, and local coverage showing the right buyer fit matter alongside the company site.

### How should I test whether ChatGPT recommends my business?

Write the questions a buyer would actually ask, such as the best service in your area for your buyer's situation, and run the same set across ChatGPT, Perplexity, and AI Overviews twice a month. Record what changes and keep the questions consistent enough to compare answers over time.

### What should a shortlist log for AI search include?

For every commercially meaningful prompt, record six fields: check date, exact prompt, engine and sign-in status, full answer saved as text, pages cited, and suppliers named, with recommendations distinguished from passing mentions. Citations go in one column and named recommendations in another, so visibility evidence stays separate from wishful thinking.

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: $19 a month flat with every feature included, and autopilot whenever you’re ready.

[Start For $19](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: it connects your AI to all your marketing and gives it proprietary marketing data collected 24/7. Plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to. groas also gives those agents large-scale proprietary marketing data it collects around the clock, on any schedule the owner chooses: every ad competitors run on Google and Meta, their landing pages and full funnels with the offers and prices in them, Reddit customer research, AI search answers from ChatGPT, Google AI Overviews, Gemini, Claude and Perplexity, and SEO data from Semrush and Ahrefs, backlinks included ($19 a month, flat, with everything included, for unlimited teammates on one shared workspace and unlimited ad accounts and websites, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#webpage","url":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your","name":"ChatGPT Cited Your Website. It Still Recommended Someone Else.","description":"A citation proves your page helped answer a question—not that your business made the shortlist. Here’s what to fix and what an AI-search provider should report.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/44d22376-dbef-4159-9018-ec7636672b48.png"},"breadcrumb":{"@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#breadcrumb"},"dateModified":"2026-11-20T09:27:58.651Z","mainEntity":{"@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"ChatGPT Cited Your Website. It Still Recommended Someone Else.","item":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#article","url":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your","mainEntityOfPage":{"@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"ChatGPT Cited Your Website. It Still Recommended Someone Else.","description":"A citation proves your page helped answer a question—not that your business made the shortlist. Here’s what to fix and what an AI-search provider should report.","datePublished":"2026-11-20T09:27:58.569Z","dateModified":"2026-11-20T09:27:58.651Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/44d22376-dbef-4159-9018-ec7636672b48.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"answer, business, chatgpt, page, cited, website, citation, shortlist","wordCount":2087,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/chatgpt-cited-your-website-why-didn-t-it-recommend-your#faq","mainEntity":[{"@type":"Question","name":"Can ChatGPT cite my website and still recommend my competitors?","acceptedAnswer":{"@type":"Answer","text":"Yes. ChatGPT can use your page as a source and link to it, then name other businesses as the options worth calling. A citation proves your content helped answer the question, not that your business made the shortlist."}},{"@type":"Question","name":"What is the difference between a citation, a mention, and a recommendation in ChatGPT answers?","acceptedAnswer":{"@type":"Answer","text":"A citation means the answer used your page as a source and linked to it. A mention means your business name appears in the answer text. A recommendation means your business is presented as an option that fits the buyer's request. They can overlap, but a source link alone is not an endorsement."}},{"@type":"Question","name":"Why does ChatGPT use my content as a source but recommend other companies?","acceptedAnswer":{"@type":"Answer","text":"Because a guide answers how a service should work, but the buyer asked who should do it. To build a shortlist, the answer needs supplier evidence: which facilities the company serves, what compliance steps it follows, what hours it covers, and who else confirms it does this work well."}},{"@type":"Question","name":"How do I check whether ChatGPT can actually access my website?","acceptedAnswer":{"@type":"Answer","text":"Check robots.txt settings for OAI-SearchBot, which serves ChatGPT search, separately from GPTBot, which is used for training, because their settings are independent. Also review server logs to see what actually gets fetched, and make sure useful information is not hidden behind JavaScript the fetcher skips."}},{"@type":"Question","name":"What information should a service page contain so AI answers can recommend the business?","acceptedAnswer":{"@type":"Answer","text":"A service page should state plainly the service area, the facilities served, hours, compliance steps, and what the company will not do. A buyer should not have to infer these facts from a general guide. Keep the useful guide, but add a commercial counterpart that answers whether the company is right for that specific buyer."}},{"@type":"Question","name":"Does ChatGPT only use my own website when deciding which businesses to recommend?","acceptedAnswer":{"@type":"Answer","text":"No. Outside sources confirm what a company's own page claims. For local search, one tracker of 200,000 non-branded prompts found business websites supplied 42% of citations, Google Business Profile 28.6%, and Yelp about 9.5%. Reviews, listings, and local coverage showing the right buyer fit matter alongside the company site."}},{"@type":"Question","name":"How should I test whether ChatGPT recommends my business?","acceptedAnswer":{"@type":"Answer","text":"Write the questions a buyer would actually ask, such as the best service in your area for your buyer's situation, and run the same set across ChatGPT, Perplexity, and AI Overviews twice a month. Record what changes and keep the questions consistent enough to compare answers over time."}},{"@type":"Question","name":"What should a shortlist log for AI search include?","acceptedAnswer":{"@type":"Answer","text":"For every commercially meaningful prompt, record six fields: check date, exact prompt, engine and sign-in status, full answer saved as text, pages cited, and suppliers named, with recommendations distinguished from passing mentions. Citations go in one column and named recommendations in another, so visibility evidence stays separate from wishful thinking."}}]}]}
```
