Citation Gaps: Finding Who AI Quotes Instead of You

A citation gap is the distance between the prompts that matter to your revenue and the sources AI systems actually quote. If Perplexity answers “best [category] for [ICP]” with three competitor URLs and a directory, that gap is your content and PR brief — not a mystery.

This spoke shows how to see what ChatGPT and peers cite, how to run a gap analysis, and how to close gaps without boiling the ocean. It supports the measurement and corroboration layers in the AEO playbook.

Lead answer: how to see what ChatGPT cites

Use the product the way buyers do, with logging:

1. Build a fixed list of prompts (your prompt panel).
2. Run each prompt in ChatGPT (with browsing/search enabled when that is the buyer path), Perplexity, and one Overview check where relevant.
3. Record every cited URL, named brand, and whether your domain appears.
4. Repeat on a schedule — single runs lie because answers vary.

There is no official “citation Search Console” that covers every chat product. Manual + spreadsheet still beats vibes. Semrush and similar suites help on the SERP/Overview side and for discovering competitor content that already ranks; they do not replace the chat panel.

Building the competitive prompt set

Start from money, not ego.

• Category recommendations
• “Best for [ICP / size / industry]”
• Head-to-head comparisons involving you or your top two rivals
• Problem/solution how-tos your sales team hears weekly
• Brand queries (“Is [You] worth it?”)

Cap the v1 panel at 25–40 prompts so you can re-run it weekly. Expand after the process works.

Running the gap analysis

Step 1 — Capture

For each prompt × product, log:

• Date / model or product version if visible
• Brands named
• URLs cited
• Your status: cited / named without link / absent / misrepresented
• Notes (outdated claim, wrong geography, etc.)

Step 2 — Aggregate

Build a URL leaderboard: which domains appear most across the panel? Build a brand leaderboard separately — sometimes you are named via a partner page you do not control.

Step 3 — Classify gaps

Gap type	Signal	Primary fix
Absence	Never cited on category prompts	Create answer-first pages; earn mentions
Substitution	Competitor how-to/docs cited	Out-write with clearer method + proof
Directory dominance	List/roundup sites win	PR to better lists; own comparison page
Stale corroboration	Old press quoted	Update or outpublish
Identity error	Wrong facts about you	Hallucination repair workflow

Step 4 — Prioritize

Score prompts by pipeline influence × winnability. A prompt that sales hears daily and where the cited URLs are thin blog posts is a better target than a prompt owned by a standards body.

Plays that close gaps

Own the missing answer page. If every citation is a “What is X?” post and you lack one, write it. Format guidance lives in Content Clusters for AI Visibility.

Beat the cited URL on compressability. Tables, steps, explicit ICP, dates. Do not merely match word count.

Earn a seat on the roundup. Digital PR aimed at the exact domains already cited is higher ROI than random guest posts. See Digital PR for Citations.

Strengthen entity + schema so when you are retrieved, you are named correctly (Entity Architecture).

Fix misrepresentation with the hallucination playbook (Avoiding Hallucinated Brand Facts).

Cadence

• Weekly: run a subset (10–15) of the panel
• Monthly: full panel + refresh URL leaderboard
• Quarterly: retire dead prompts, add new buyer language from sales calls

Share a one-page dashboard with leadership: citation rate, share of voice vs named set, top gap URLs, shipped fixes.

Checklist

• Prompt panel documented and owned
• Logging template in use (sheet or Notion)
• Competitor set frozen for 90 days (avoid moving goalposts)
• Top 10 cited external URLs reviewed manually
• Three gap-closing pages or placements queued
• Re-measure date booked

Spreadsheet columns that keep teams honest

Use boring columns:

date | product | prompt_id | prompt_text | brands_named | urls_cited | our_status | accuracy | notes | owner

Status enum: cited_link | named_only | absent | misrepresented.

Accuracy enum: ok | minor | material.

Export a pivot monthly: prompts where our_status=absent and competitors appear ≥2 times. That pivot is the content/PR backlog.

Competitive set discipline

Freeze 3–6 competitors for 90 days. If sales adds a new rival mid-quarter, park them in a “watch” list rather than reshaping every chart. Moving goalposts hide whether your work worked.

Include one “aspirational” competitor that currently owns citations even if they are larger. Include one peer. Include one cheap substitute. Different gap plays apply to each.

Qualitative review of top cited URLs

For each of the top 10 external URLs:

• What format is the citeable passage? (list, table, definition, docs)
• How fresh is the page?
• Are facts about the category accurate?
• Is there a path to earn inclusion (contribute, earn roundup slot, outpublish)?
• Would we be proud to be quoted next to them?

If the winning URL is a thin listicle, outpublishing is viable. If it is a primary standards doc, change the prompt strategy — do not cosplay as ISO.

When absence is correct

Sometimes AI omits you because you are out of scope for the prompt (“best enterprise suite” when you serve mid-market). Do not force inclusion with spam. Either:

• Narrow your prompt panel to ICP-true questions, or
• Expand the offer honestly if the market shift is real

False inclusion that creates bad-fit demos is not a win.

Sharing gaps with executives

Lead with money prompts. Show one side-by-side: competitor cited URL vs your missing page. Ask for budget against the gap list, not against “AI” as a vibe. Revisit in 60 days with the same prompts.

Gap severity scoring

Score each absent money prompt 1–5 on:

• Pipeline influence (sales hears it)
• Frequency in buyer research
• Winnability (cited URLs look beatable)
• Strategic fit (we actually want those leads)

Multiply influence × winnability for a rough priority. High influence / low winnability items become long-term PR plays, not weekend blog posts.

Stealing structure, not sentences

When a competitor URL dominates, reverse-engineer structure: their H2s, table shape, and FAQ list. Then beat them on:

• Clearer ICP boundaries
• Fresher dates
• Original proof
• Better internal links to your offer
• Accurate entity markup

Do not paraphrase their prose. Retrieval systems and humans both notice derivative sameness — and your lawyers might too.

Multi-product citation matrices

Build a matrix with prompts on rows and products (ChatGPT / Perplexity / Overview) on columns. Color cells green/yellow/red for your status. Executives grasp matrices faster than paragraphs. Update monthly screenshots into the same slide so trends are obvious.

Partner and marketplace gaps

Sometimes you are absent because buyers ask inside ecosystems (Shopify app stores, AWS marketplace, HubSpot directories). Those surfaces need their own listings and descriptions aligned to the fact packet. Classic web PR will not fix a thin marketplace listing that AI mirrors.

Add ecosystem prompts to the panel when a material share of pipeline starts there.

Closing the loop with content ops

Every “absent” cell should become a ticket with type page | pr | listing | entity | wontfix. “Wontfix” is allowed when the prompt is out of ICP. Track wontfix reasons so sales does not keep asking why you ignore vanity queries.

Implementation notes: weekly gap standup

Run a 25-minute standup:

• Three minutes: citation rate sparkline
• Seven minutes: review new absent money prompts
• Seven minutes: assign or advance tickets (page / PR / listing)
• Five minutes: accuracy incidents
• Three minutes: next retest commitments

Invite content, PR, and someone from sales at least twice a month so prompt language stays real. Keep the meeting sacred — when it slips, gap analysis becomes a quarterly archaeology dig.

Store decisions in the sheet, not in chat. Six weeks later nobody remembers why you marked a prompt “wontfix,” and the argument restarts.

Example gap narrative for leadership

“On 12 recommendation prompts for our ICP, we appear in 2. Competitor A appears in 9, mostly via two roundup URLs and their comparison page. We lack an equivalent comparison page and are absent from both roundups. Proposed: ship comparison page in two weeks; pitch both editors with updated boilerplate and a data point from last quarter’s delivery metrics. Retest at day 45.”

That paragraph is more fundable than a generic AI strategy deck.

Practical week-one kit

Create four artifacts on day one of a gap project: the prompt panel sheet, the competitor freeze list, the URL leaderboard pivot, and a backlog tagged page/PR/listing/entity/wontfix. Until those four exist, you are still in opinion mode. Once they exist, every meeting can point at a cell and a ticket. That is how citation gap analysis becomes boring enough to run forever — which is the point.

Repeat the kit after major launches. The cost of re-baselining is tiny compared with a quarter of unmeasured content. Keep owners named in the sheet. When someone goes on leave, transfer the ritual explicitly — AEO dies in the handoff gaps. If you need a second pair of eyes, the visibility lane exists for that reason: /visibility and the visibility audit path turn these kits into a managed baseline with a 30/60/90 plan. Either way, ship the ritual before you buy another dashboard logo.

FAQ

How do I see what ChatGPT cites?

Enable browsing/search when relevant, ask the buyer-style question, and record the citation list. Repeat across sessions and pair with Perplexity and Overview checks. No single export replaces that panel today.

What is an AI citation gap analysis?

It is a structured comparison of the sources and brands AI systems cite for your priority prompts versus your own presence — then a prioritized plan to earn inclusion.

How many prompts do I need?

Enough to cover revenue-relevant intents without drowning the team. Most brands can learn from 25–40 well-chosen prompts re-run weekly.

Why do directories keep winning citations?

They compress category advice into lists models can quote. Either earn placement on quality lists, publish a better-owned comparison with clear criteria, or both.

Can Semrush show ChatGPT citations?

Use Semrush for SERP, Overview, and competitive content discovery. Treat chat citations as a separate logging workflow unless a tool you trust proves product-level coverage you can verify.

How fast can gaps close?

On-site pages can enter browsing-mode answers within days to weeks. Displacing entrenched roundups may take sustained PR. Measure monthly, not daily.

Closing

Stop guessing who AI quotes. Log it, classify the gap, ship the fix, re-measure. That loop is competitive AEO.

For the surrounding system, read the AEO playbook. Spurlock Studios runs citation-gap baselines inside visibility audits — start at /visibility or book an audit.