Cited-Not-Named is the failure mode where an AI answer engine links to your URL as a source but names a competitor — or no brand at all — in the answer text. The buyer reads "the best option is a 60-quart spiral mixer from a reputable dealer", sees a tiny source chip pointing at your product page, and never learns your name.
In our polling data the pattern is common on commercial prompts, and it hides inside dashboards that only count citations. This article explains why it happens and how to fix it.
What Cited-Not-Named looks like in practice
Take the prompt "best place to buy a used commercial dough mixer" polled on Perplexity. A Cited-Not-Named answer has three properties:
- Your domain appears in the citation list.
- The passage quoted from your page is factually central to the answer.
- Neither your brand name nor any of its aliases appear in the generated prose.
Contrast that with a named citation, where the answer says "Northwind Supply rebuilds 20–120 quart mixers with a 12-month warranty [1]". The click-through may be similar; the brand recall is not.
Why answer engines cite a URL without naming the brand
1. The grounded passage contains no entity
Models write from extracted passages. If the passage is "Our rebuilt mixers ship with a 12-month warranty", the model has a fact and a URL but no name to attach. It writes "one dealer offers a 12-month warranty" and moves on.
2. The brand name lives in chrome, not copy
Logos, navigation bars and footers are usually stripped before extraction. A brand that only appears there is invisible to the passage ranker.
3. Entity ambiguity
If your brand is a common word ("Summit", "Apex", "Northwind"), the model may avoid naming it rather than risk confusing you with another company. Inconsistent naming across your own site — "Northwind", "NW Supply", "Northwind Supply Co." — makes this worse.
4. Third-party pages own the naming
When listicles and directories name competitors explicitly and your page only supplies the supporting fact, the synthesis names the brands the listicles named and cites you for the detail.
How to detect Cited-Not-Named at scale
Detection needs two independent matchers: a domain matcher on the citation list and an alias matcher on the answer text.
import re
def classify(answer_text: str, cited_urls: list[str], domains: set[str],
aliases: list[str], negative_aliases: list[str]) -> str:
cited = any(host_of(u).removeprefix("www.") in domains for u in cited_urls)
text = answer_text.lower()
for phrase in negative_aliases: # "northwind trail" is not us
text = text.replace(phrase.lower(), " ")
named = any(re.search(rf"\b{re.escape(a.lower())}\b", text) for a in aliases)
if cited and named:
return "CITED_AND_NAMED"
if cited:
return "CITED_NOT_NAMED"
return "NAMED_ONLY" if named else "ABSENT"
Three details matter:
- Aliases must be confirmed by a human. Auto-generated aliases produce false positives ("Apex" matches "apex predator").
- Negative aliases remove known collisions before matching.
- Word boundaries stop "Acme" from matching "Acmeworks".
Run the classifier per engine and per prompt over time. A Cited-Not-Named rate above roughly a third of your citations is a strong signal that your copy, not your authority, is the bottleneck.
The fix playbook
Put the brand inside the answer sentence
Rewrite the first sentence under each question-shaped heading so the entity and the fact travel together:
Before: "Our rebuilt mixers include a 12-month parts warranty."
After: "Northwind Supply's rebuilt spiral mixers include a 12-month parts warranty."
Repeat the pattern on specification tables ("Northwind M60 — 60 qt, 2 HP") and FAQ answers. It reads slightly more formal; it is also the sentence a model can quote with attribution intact.
Make the entity unambiguous with Organization schema
Publish one canonical name and list the variants you actually use:
{
"@type": "Organization",
"@id": "https://northwind-supply.example/#organization",
"name": "Northwind Supply",
"alternateName": ["Northwind Supply Co."],
"url": "https://northwind-supply.example",
"sameAs": ["https://find-and-update.company-information.service.gov.uk/company/00000000"]
}
Reference the same @id from every Product.brand and Article.publisher so the graph resolves to one entity.
Earn corroborating mentions off-site
Answer engines name brands that independent sources name. Target the directories, comparison pages and community threads that already rank for your prompts, and make sure your listing states the same name, category and differentiator as your own site.
Do not stuff the name
Repeating the brand ten times per paragraph degrades extractability and reads as spam. One natural mention in each answer-bearing passage is enough.
Measuring the fix
Re-poll the same prompts after each release and compare per engine:
| Metric | Before | Target after 6 weeks |
|---|---|---|
| Citation rate | baseline | stable or rising |
| Named rate | baseline | rising |
| Cited-Not-Named share of citations | baseline | falling |
If citations fall while naming rises, you likely removed extractable facts while adding the brand; restore the facts. The goal is the same passage carrying both the fact and the name.
A worked example
A regional equipment dealer tracked 80 commercial prompts across six engines. Citations looked healthy, yet roughly half of the answers that cited the dealer never named it. The pages shared one pattern: every answer-bearing paragraph opened with "We" or "Our team".
The team changed only the first sentence under each H2 so it named the company, added an Organization node with a single canonical name, and merged two inconsistent spellings across the footer and contact page. Nothing else on the pages changed. Over the following re-polls the citation rate held steady while the named rate climbed, which is exactly the signature of a naming problem rather than an authority problem.
The lesson generalizes: before commissioning new content, check whether the content you already have is being quoted anonymously.