Across roughly 16 million brand appearances, new data from Writesonic found that around 40% of AI citations didn't name the source brand in the generated answer. On Perplexity, the rate reached 52%.
(Disclosure: the article co-author, Samanyou Garg, is the founder and CEO of Writesonic.)
Getting cited by an AI engine is often touted as a win. You did the work, your page appears in the sources, your URL shows up in citation reporting.
But one thing people in our field aren't often talking about is the fact that even with a citation, the reader may never see your brand name.
We call this a ghost citation: an AI engine links to your page as a source but doesn't mention your brand in the answer itself.
That distinction changes what a citation actually tells you. It confirms that your content contributed to the answer, but not that your brand was visible to the reader.
A citation means an AI answer links to your page as a source — inline, as a footnote, or in a sources panel. A mention means your brand name appears in the generated answer text.
With a citation and a mention:
"Writesonic's analysis of roughly 16 million brand appearances found that approximately 40% of AI citations didn't name the source brand."
A ghost citation:
"One analysis found that approximately 40% of AI citations didn't name the source brand."
Both answers could link to the same Writesonic page. But only the first one tells the reader who produced the research.
That is the ghost citation gap: your content is visible to the engine, while your brand may remain invisible to the reader.
| AI engine | Ghost citation rate |
|---|---|
| Perplexity | 52% |
| Google AI Mode | 49% |
| Google AI Overviews | 41% |
| ChatGPT | 37% |
| Gemini | 25% |
| Grok | 22% |
| Microsoft Copilot | 19% |
| Overall | ~40% |
This means two brands can earn the same number of citations and receive very different levels of recognition depending on the engine.
Out of 100 cited appearances on Perplexity, the brand would be absent from the answer text in approximately 52. On Copilot, approximately 19. The citation count could look identical in a report, but the number of times the reader actually sees the brand wouldn't.
Neither outcome is ideal. A mention without a citation makes the brand visible but may not provide the supporting source and encourage referral clicks. A citation without a mention makes the page available but may leave the brand out of the response, arguably missing the more important outcome: brand recognition. The strongest outcome is receiving both.
This split means optimization strategy should differ by engine. Chasing more citations on Perplexity may increase your source count without increasing how often your brand appears in the answer. On Gemini or Copilot you're more likely to be named, but need to be cited more consistently to drive referral traffic.
Ghost citations are likely the result of how different AI engines process and synthesize sources rather than any deliberate choice to omit brand names. But the honest answer is: our field doesn't have a complete explanation yet.
One possible explanation is that some engines treat the linked citation as sufficient attribution and then present the information in their own "voice." Google's AI answer offers similar tips to its citations but its style, tone and phrasing are unique — the engine has absorbed the source materials and rewritten them, leaving the citation in the footnotes but the brand name out of the answer.
Content structure may also play a role. When a brand name isn't closely paired with the claim it produced on the source page, synthesis has more room to sever them.
A statistic buried three paragraphs below the company name that generated it is probably easier to anonymize than one where attribution is built into the sentence itself. This is not a proven rule, but something I'd suggest testing.
Classify tracked appearances into four groups:
Then segment by engine — an aggregated citation score can hide large differences between platforms.
Proprietary research or data can become your brand moat because it may be harder to ghost than generic advice.
When a study, index, framework or methodology carries your brand's name, the data and the brand become harder to separate because the finding references you by definition.
Named research also gives journalists a stronger hook, and that earned coverage matters doubly now: third-party content gets cited more heavily by AI engines, so the offsite PR your named research generates compounds into more AI citations over time.
Instead of:
❌ "Our analysis found that around 40% of AI citations didn't name the source brand."
Use:
✅ "Writesonic's Ghost Citation Study found that around 40% of AI citations didn't name the source brand."
That association shouldn't live only on your own site. Independent coverage, expert commentary, customer content, and industry discussions across the sources that most influence AI answers can repeatedly connect your brand with the ideas it wants to own.
For important prompts where your page is cited but your brand is missing, review:
Then test whether clearer attribution, consistently named research, or page structure changes affect the result over time.
A source link confirms that an AI engine used your content. It doesn't confirm that the reader saw your brand. That's why citation rate can't be the only measure of AI visibility.
Citation rate tells you that an engine found your content. Mention rate tells you whether your brand actually traveled with it. A citation your reader never sees is visibility that only exists in a dashboard.
Writesonic analyzed a recent 30-day window of AI answer data across thousands of brands, covering approximately 16 million brand appearances across Perplexity, Google AI Mode, Google AI Overviews, ChatGPT, Gemini, Grok and Microsoft Copilot.
Percentages were rounded. The results represent a snapshot of engine behavior during the measurement window and don't establish the causes of ghost citations or their effect on clicks, brand recall, trust, or conversions.
Retire citation rate as a single metric. The article's core claim: citation and mention are separate outcomes and must be measured separately.
Set different strategies per engine. 52% on Perplexity versus 19% on Copilot is not a gap one strategy can address.
Put the brand inside the sentence. The most immediately actionable item — swapping "our analysis found" for "[Brand]'s [Named Study] found" reduces the room synthesis has to sever them (hypothesis-stage, but near-zero cost).
Name your research. Named studies, indexes and frameworks bind data to brand by definition — the only structural moat this article offers.
Don't claim to know the cause. The authors state plainly that the field lacks a complete explanation. A vendor asserting a definitive cause is the first thing to question.
Plan it with earned media. That third-party content gets cited more heavily is the basis for the named-research strategy — see 82% of AI Citations Point to Earned Media.
Fold it into your measurement framework — see Do We Show Up in ChatGPT? and, on whether the brand registers as an entity at all, Entity SEO.
Check retrievability first. Not being named may be because the brand entity itself is unclear — see AI Search Can't Verify Your Business.
When an AI engine links to your page as a source but doesn't mention your brand in the answer text. It confirms your content contributed to the answer, but not that your brand was visible to the reader.
Perplexity 52%, Google AI Mode 49%, Google AI Overviews 41%, ChatGPT 37%, Gemini 25%, Grok 22% and Microsoft Copilot 19%, with an overall average around 40%.
Namers like Gemini and Copilot include brands in answer text more often but link less frequently. Citers like Perplexity and Google's answer engines link more often but name brands less consistently. ChatGPT and Grok sit between them.
Measure citations and mentions as four separate groups segmented by engine, build named research your brand owns, pair the brand name directly with the claim in the sentence, and investigate individual answers where you were cited but not named.
Writesonic analysed a recent 30-day window of AI answer data across thousands of brands, covering approximately 16 million brand appearances. Results are a snapshot of engine behaviour and do not establish causes or effects on clicks, recall or conversions.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
An E-E-A-T checker was never automatable because there is no API for it. Here is how an AI coding agent read Google's quality guidance, crawled a page sample, and produced a scored audit — plus where it stops.
A GEO audit is not checking whether your brand appears in an AI answer. Here is a nine-point GEO checklist split into four on-site items, four off-site, and one measurement layer.
LinkedIn, YouTube, Reddit, and TikTok are building systems to identify low-value AI content. Here are the limits of watermarking and what content still gets distributed.