What You'll Learn

A SEMrush study published June 9, 2026 gave the marketing world a new term: ghost citations. The AI links your site as a source. Your brand name never appears in the answer. Sixty-two percent of AI citations work this way — you get the footnote, not the mention.

But there's a worse outcome the study doesn't cover. That's the one actually costing brands deals.

Ghost Citations: When AI Uses Your Content but Forgets Your Name

The study analyzed 3,981 domain appearances across 115 prompts, 14 countries, and four AI search engines. The breakdown: 61.7% ghost citations (cited, not named), 13.2% both cited and named, and 25.1% named without any citation. Nearly two-thirds of all AI source links carry no brand mention — the site earns the reference, the buyer never sees who made it.

ChatGPT cites sources 87% of the time but names brands in only 20.7% of appearances. Gemini nearly inverts it: 83.7% mention rate, 21.4% citation rate. Citation behavior and mention behavior are almost entirely decoupled — and vary dramatically by engine. Co-author Kevin Indig: "There's almost no overlap between which brands ChatGPT cites and which ones Gemini names for the same prompt. These are different behavioral systems. Treat them that way."

61.7%
of AI citations are ghost citations — your site earns the source link, your brand name never appears in the answer. SEMrush, June 2026
87% → 20.7%
ChatGPT's citation rate versus its brand mention rate on the same appearances. Being a cited source and being a named brand are two separate outcomes. SEMrush, June 2026

Ghost citations are addressable — optimize for AI extractability and your name starts surfacing alongside the link. What comes next on the spectrum is not so clean.

Worse Than a Ghost Citation: When AI Mentions You — and Gets It Wrong

The worst outcome isn't invisibility. It's a confident, cited, wrong description of your brand delivered to a buyer who has no reason to doubt it.

seoClarity tracked 8 months of branded prompts and found 40% of AI brand responses contain an inaccuracy — wrong capability, outdated positioning, a feature attributed to the wrong product, all dressed up with a citation that makes it look authoritative. 12% of branded ChatGPT queries include a factually incorrect claim, and 35% of brands say AI inaccuracy has already damaged their reputation. The disqualification never appears in your analytics — the buyer reads the wrong story and moves on.

Real Example — Seer Interactive

Seer Interactive is a 24-year-old marketing agency. In 24 years, they received one negative review — posted in 2018, then duplicated across review aggregation sites until a single data point looked like a pattern. It appeared in at least 1 in 3 branded AI prompts, and on three of five review sites cited, it was the only review listed.

AI presented it as representative sentiment — not an outlier, but the defining description of the company. The brand was named. The source was cited. The story was wrong. That's a ghost narrative, and it's the outcome that costs deals.

BrightEdge AI Catalyst found Google AI Overviews are 44% more likely to criticize brands than ChatGPT, and ChatGPT is 13x more likely to go negative at the point of purchase. G2's 2026 research found 69% of B2B buyers chose a different vendor than they planned based on AI guidance. When AI names you and gets the story wrong, you're not losing a click. You're losing a deal.

"A ghost citation is when AI cites your site but never says your name. A negative mention with citation is when AI names you, says something wrong, and links to a source that makes it look credible. The first costs you visibility. The second costs you deals. Shensuo finds both."

How to Fix Ghost Citations and Negative AI Mentions Before They Cost You Deals

Both problems are fixable. Neither is visible without the right data.

Ghost citations are an SEO gap between your content and your brand identity. The fix is concrete: identify which queries are ghost-citing you, then strengthen the entity signals that prompt AI to name you alongside the source link. Most brands don't have that list. They're optimizing blind.

Negative mentions with citations require a different approach. You can't correct the content feeding a bad narrative without knowing what narrative AI is producing. That description is invisible in your rank tracker and your traffic data. Shensuo surfaces both:

What You Actually Need to Measure
  1. Ghost citation identification. Which queries are sourcing your site without naming your brand — broken down by model, query type, and buyer intent stage. That's the input list for fixing your content so citations convert into mentions.
  2. Brand story scoring. The exact sentences AI uses to describe you when it does say your name — flagged by sentiment and accuracy. Wrong pricing, outdated features, borrowed competitor language, stale positioning. This is what feeds the fix: you see the bad output, you trace it to the source content, you correct it.
  3. Cross-model narrative comparison. What ChatGPT says about your brand versus Gemini versus Perplexity, side by side, for the same prompt. BrightEdge data shows these platforms disagree on brand narratives 62% of the time. Your brand doesn't have one AI story — it has three, diverging in ways that matter at evaluation stage.

SparkToro research found less than a 1-in-100 chance ChatGPT gives the same brand list across any two sessions. A single brand audit tells you what one buyer saw, once. You need systematic scanning across sessions, models, and query variants to see where you're ghost-cited and where the narrative breaks when you are named.

When HubSpot launched its AEO product in April 2026, it settled the measurement question: AI answer visibility is a required metric. The remaining question is whether yours is shallow — did my name appear? — or deep enough to catch both the silence and the story. Most brands are flying blind on both. Shensuo shows you both.