# Brand mentions versus citations: what is the difference, and which is worth more?

> Your name appearing in an AI answer and your site being cited are separate events. How they differ, how they vary by engine and how to read them in a report.

Published: 2026-10-07
<https://geomono.com/en/blog/marka-anilmasi-ve-kaynak-gosterilmesi-farki>

A brand mention is your brand's name appearing in the text of an AI answer. A citation is the engine, while building the answer, listing your site among the links the answer relies on. The two can occur together in the same answer, or one can occur without the other; we know of no data that measures which is worth more. This article explains why the two need to be measured separately and what each one is good for.

## Two separate events: mentions and citations

A mention concerns the text the reader sees. If your name appears as an option in the answer to "İstanbul'da kurumsal müşterilere hizmet veren bir bordro yazılımı önerir misin?" ("Can you recommend a payroll software that serves corporate clients in Istanbul?"), the person reading the answer has heard of you. Even if they click no link at all, your name may stay with them.

A citation concerns what the engine relies on. The engine searches the web, reads some pages and shows them as links beside or below the answer. Those links may go to your site, to a marketplace, to a complaints site or to a news article.

Engines do not show sources in the same way, either. According to [Perplexity's help page](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work), every answer includes numbered links to the original sources. [OpenAI's help page](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt), on the other hand, writes that ChatGPT answers that use web search may include source links; in other words, not every answer. Google AI Overviews and AI Mode show supporting links beside the answer, and according to [Google's documentation](https://developers.google.com/search/docs/appearance/ai-features), for a page to be one of those links it must be indexed and eligible to be shown with a search result snippet.

Combining the two events gives four situations:

| Situation | Your name in the answer text | Your site as a source | What it means |
|---|---|---|---|
| Mentioned and cited | Yes | Yes | The engine read your site and told the reader your name |
| Mentioned, not cited | Yes | No | Your name came from other pages or from the model's own knowledge |
| Cited, not mentioned | No | Yes | The engine read your site but did not tell the reader your name |
| Neither | No | No | You did not appear for this question |

The fourth row is valid only if the answer was actually measured. If the engine returned an error or the source list could not be retrieved, that answer does not enter this table at all. "Could not be measured" is one situation, "absent" is another, and the two must not be recorded as the same zero.

## Read, but not named: our own measurement

The third row is more common than you might think. In 60 analyses we ran up to 3 August 2026 for our own client and comparison brands, we counted 2,748 answers that cited the brand's own site. In 341 of them (12.4%), the brand name did not appear anywhere in the answer text.

We recounted these figures from the same records on 7 October 2026. The first count had found 2,733 answers and 12.8%; corrections made later to the stored data changed the count slightly. The recount did not change the direction of the finding: in roughly one answer in eight, the site had been read but the name had not been said.

The limits of this measurement need to be understood:

- **Who was measured:** 19 company records, that is, our own client and comparison brands. They are mostly technology and consumer brands; this is not a sample representative of the Turkish market.
- **Which questions:** Both questions that contain the brand name and questions that do not; mostly product and category questions.
- **Which engines:** ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Google AI Mode was not part of this measurement.
- **How "name not present" was determined:** By automatic name matching on the answer text. Namesakes and different spellings may affect the result.

In the same measurement, most source links went outside the brand's own site anyway. Of the 75,914 source links captured, 4,887 went to the brand's own site and 71,027 to other sites (93.6%). In 6 of the 60 analyses, the brand's own site was never cited. That is why the question "is our site being cited?" is not enough on its own; you also need to see the other pages that describe you.

## The difference by engine

The "cited but not mentioned" rate varies widely from engine to engine. In the same measurement, the answers that cited the brand's own site broke down by engine as follows:

| Engine | Answers citing the brand's own site | Brand name not present | Rate |
|---|---|---|---|
| ChatGPT | 813 | 40 | 4.9% |
| Gemini | 525 | 20 | 3.8% |
| Perplexity | 793 | 155 | 19.5% |
| Claude | 515 | 118 | 22.9% |
| Google AI Overviews | 102 | 8 | 7.8% |

This table is not here to rank the engines. The Google AI Overviews row rests on only 102 answers; on a base that small, a single analysis can move the rate noticeably. What the difference says is simpler: being cited does not turn into being mentioned to the same degree on every engine.

The habit of citing sources itself also varies by engine. In a separate measurement, between 16 and 25 August 2026, we asked only questions without the brand name for three corporate brands (a pharmaceutical company, a dairy brand and a technology group). Across the 1,646 answers measured, the share that cited at least one source was as follows by engine:

| Engine | Answers measured | Citing at least one source | Rate |
|---|---|---|---|
| ChatGPT | 362 | 180 | 49.7% |
| Gemini | 341 | 321 | 94.1% |
| Perplexity | 339 | 339 | 100% |
| Claude | 322 | 269 | 83.5% |
| Google AI Overviews | 282 | 228 | 80.9% |

In this measurement, 189 (10.3%) of the 1,835 planned answers could not be measured and were excluded from the rates: 166 returned an error, for 16 the source information could not be retrieved, and for 7 there was no AI Overviews block to read. Because the two measurements used different brands, different questions and a different method, their percentages should not be compared with each other.

About half of the answers measured on ChatGPT cited no source at all. Your citation count on that engine may come out low, but that does not mean your name is absent from those answers. On Perplexity, by contrast, every answer measured cited at least one source; there the source list is always present, but being on the list does not guarantee that your name appears. Combining the two indicators into a single number erases this difference.

{{cta}}

## Which is worth more?

The honest answer: we have no data that measures this. To know whether mentions or citations bring in more customers, you would have to track what the people reading the answer do afterwards. Our own measurements record what the engines write and what they cite; they do not record what the reader does after the answer.

So we do not write sentences like "a mention is worth this much more than a citation". When you see such a sentence, we suggest asking which data, how many answers and which engine it rests on. Even without a measured ranking of value, what each one is good for is clear.

### What is a mention good for?

A mention is where the reader sees the options. When a buyer asks "hangi bordro yazılımını seçmeliyim?" ("which payroll software should I choose?") and your name is not in the answer, it is hard for that person to see you among the options, even if your site is in the source list. A mention also shows how you are mentioned: as the leading option, as one of the alternatives, or with a caveat.

### What is a citation good for?

A citation shows what the engine relies on. If your site is cited, the engine can reach and read your site; that tells you a technical precondition is met. The link also opens a path for the reader to your site. If the source list contains a complaints site or a marketplace rather than your site, you learn through whose account the engine knows you.

### What does "read but not mentioned" tell you?

This situation shows the gap between the two. The engine reached your site but did not tell the reader your name. We cannot know why from a single number; the answer and the cited page need to be read side by side. If the page answers the question but does not clearly carry the brand name, that is the first place to look. Read this as a lead to investigate, not as a guarantee.

You can check whether your site is open to AI crawlers and whether your pages are citable with our [scan, which requires no sign-up](/en/tarama). The scan looks at the technical precondition for being cited; it does not measure whether you are mentioned.

## Showing the two separately in a report

A good report does not add mentions and citations into one number. Each is written with its own pair of numbers and separately for each engine. The numbers below are examples given only to show the format:

- **Mention:** "Mentioned in 12 of 40 questions asked without the brand name" (example).
- **Citation:** "Your site was listed in 7 of 30 answers that cited at least one source" (example).
- **Overlap:** In how many of the answers that cited your site your name did not appear.
- **Unmeasurable:** Answers that returned an error or whose source list could not be retrieved, on a separate line; they enter no rate as a zero.

The report should also state what the base is. If half of the answers measured on ChatGPT cite no source at all, whether you calculate the citation rate over all answers or only over answers that cite a source changes the result a great deal. Both are defensible; not stating which one you used is not.

On the source side, looking only at your own site also falls short. Our [Sources and Competitors](/en/urun/kaynaklar-ve-rakipler) screen collects, by domain, which third-party pages the engines read in the answers where your name does not appear, and notes next to each which engine cited which source. Answers whose source list could not be collected are counted separately there too, not written up as "no sources". That appearing in a given source would change the answer has not been proven, and the page says so explicitly.

For how to write the questions, why the engines are asked separately and how unmeasurable answers are set apart, see the guide [how to measure brand visibility in AI](/en/blog/yapay-zeka-marka-gorunurlugu-nasil-olculur).

> [!UYARI]
> When you see a number that says "your citation rate is this", ask three things: on which engine, over how many answers, and were unmeasurable answers mixed into this number as zeros?

## Frequently asked questions

### If my site is cited, does my brand count as visible?

Not on its own. In our own measurement, the brand name did not appear in the answer text in 341 (12.4%) of the 2,748 answers that cited the brand's own site. Being cited shows that the engine read your site; it does not show that the reader saw your name.

### Which is worth more, a mention or a citation?

We know of no data that measures this; our own measurements do not record what the reader does after the answer. A mention shows where the reader sees the options, while a citation shows what the engine relies on. Tracking the two separately gives more information than substituting one for the other.

### Why is this difference larger on some engines?

Engines show sources in different ways and with different frequency. In our own measurement, the "cited but not named" rate came out at 22.9% on Claude and 3.8% on Gemini. This measurement is limited to our own client and comparison brands, and the reason for the difference cannot be stated without reading individual answers; that is why each engine should be reported separately.

### If my name never appears, what should I do first?

First make sure this was actually measured: answers that returned an error or whose source list could not be retrieved must not be counted as "not present". Then read the answers that cited your site but did not name you side by side with the page that was read in those answers. Checking whether the engines can reach your site is also one of the first steps.

## Sources

- [How does Perplexity work?](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work), Perplexity Help Center, updated 3 September 2026
- [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt), OpenAI Help Center, accessed 7 October 2026
- [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features), Google Search Central, updated 10 December 2025
- Answers that cited the brand's own site but did not name the brand, 2026-08-03: 60 analyses up to 3 August 2026, 19 company records (our own client and comparison brands, not a market sample), ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews; 2,748 answers that cited the brand's own site, recounted on 7 October 2026 (first count 2,733); name matching is automatic, namesakes and spelling variants may affect it
- How often engines cite sources, questions without the brand name, 2026-08-25: 6 measurements between 16 and 25 August 2026, 3 corporate brands, five engines; 1,646 of 1,835 planned answers measured, 189 (10.3%) could not be measured and were excluded from the rates