Insight · 10 min
Four patterns engines use — remediation when you're named but not trusted.
AI answers use two different signals that SEO teams often collapse into one:
You can have high mentions and weak citations. You can be cited while a competitor gets the recommendation. Treat them as separate metrics with separate remediation playbooks.
The model includes your brand, product name, or unmistakable entity reference in the response text.
Types MentionX extracts:
| Type | Example |
|---|---|
| Explicit | “Consider VendorCo for mid-market teams…” |
| Implicit / product | “Tools like VendorCo Pro support…” |
| Absent | Category discussed; your brand never appears |
Metric: mention rate = mentions / engine runs on buyer prompts.
A URL, domain, or structured source reference attached to the answer — common in Perplexity, increasingly in grounded Gemini and web-enabled ChatGPT.
Types:
| Type | Example |
|---|---|
| Direct link | Footnote to yourdomain.com/pricing |
| Inline source list | Numbered references at end of answer |
| Third-party | G2, Reddit, Wikipedia, analyst report |
Metric: citation share = your citation credit / sum of citation credit for you + tracked competitors (not “all URLs in the scan”). See dashboard metrics explained.
| Pattern | Mention | Citation | What it means |
|---|---|---|---|
| A — Ideal | ✓ | ✓ Your domain | Strong entity + owned proof |
| B — Silent authority | ✗ | ✓ Your content | Your docs/blog support category answer but brand not named — positioning gap |
| C — Hollow visibility | ✓ | ✗ or third-party only | Named from model memory; no path to your proof — fragile vs updates |
| D — Invisible | ✗ | ✗ | Not in consideration set |
Run prompts. Classify which pattern you are in per engine — they diverge.
Buyers act on names in the shortlist. A mention puts you in the consideration set even when zero-click.
Measure:
If mention rate is 0.15 and a competitor is 0.70 on identical prompts, the pipeline problem precedes traffic analytics.
Citations show what evidence the engine trusted when it formed the answer.
Repeated citation domains in your category often include:
Citation gap (MentionX terminology): a domain appears in ≥N competitor citations and zero yours on the same scan — a source-type you do not participate in.
Fixing citation gaps is not “more blog posts.” It is strategic presence in the source types engines already cite for your competitors.
| Engine | Mention behavior | Citation behavior |
|---|---|---|
| ChatGPT (web-enabled) | Strong shortlists from retrieval + training | Citations when browsing; not always to your domain |
| Gemini (grounded) | Google-retrieval influenced | Often cites pages Google trusts |
| Perplexity | Answer + numbered sources | Citations explicit — easiest to audit |
Implication: A win on Perplexity citations does not automatically transfer to ChatGPT mentions. Measure separately.
Step 1 — Run a fixed prompt set
10–50 buyer intents (recommendation, comparison, alternative).
Step 2 — Tag each response
Mention? Position? Citations (domain list)?
Step 3 — Build a citation domain matrix
| Domain type | Competitor A cites | You cite | Action |
|---|---|---|---|
| g2.com | 12 | 0 | Review profile + category presence |
| reddit.com | 8 | 1 | Community threads / AMA / use-case posts |
| yourdomain.com | 2 | 5 | Good — expand comparison URLs |
| wikipedia.org | 1 | 0 | Entity notability / disambiguation audit |
Step 4 — Prioritize by prompt intent
Fix comparison prompts before vanity “brand awareness” prompts.
MentionX stores every engine response and extracts:
Formulas: MentionX Methodology.
Are citations the new backlinks?
Similar in importance, different in mechanism. Citations in AI answers are evidence selections, not PageRank transfers. Measure them directly.
Should we optimize for Perplexity first?
Perplexity is the easiest citation audit. Use it as a diagnostic, not your only channel.
Do mentions require citations to convert?
Buyers may act on mentions alone (zero-click). Citations increase trust and durability when models refresh.
