AI citation gap analysis: manual owned-link evidence
This guide defines a manual AI owned-link gap as an eligible completed prompt-surface-repeat cell in which the target site lacks a qualifying UI-marked citation or supporting-source link under a frozen role and host rule. It is not RankEcho's product citation-gap state. Register the prompt, service, surface, repeat, locale, account state, model label when exposed, and collection time before classifying the cell. Preserve unavailable and failed cells, the raw answer, and every visible URL. Compare source coverage without assigning cause, then choose the smallest check that can distinguish among plausible explanations. Move to Fix only when that check establishes a repairable defect. Site traffic is a separate outcome and is not required to complete this analysis.
What is the scope of an AI citation gap analysis?
Use this analysis to describe one bounded absence of a visible target-owned source in a recorded answer. The unit is not a keyword, domain, dashboard total, or anecdotal chat. It is one registered prompt, service and surface, locale, account state, model label when available, collection time, and citation-matching rule.
The analysis separates what the answer displayed from hypotheses about access, entity resolution, prompt fit, source coverage, or answer variability. It does not identify a private selection mechanism, promise a future citation, or convert low site traffic into evidence of a citation problem.
How does this manual gap differ from the RankEcho product metric?
The current RankEcho product scorecard rolls up by prompt and can mark its headline citation state from a target URL or a detected brand or domain mention. This manual worksheet counts prompt-surface-repeat cells and requires a qualifying target-owned source link; a mention-only observation remains negative.
Do not use the product scorecard or Proof Loop to recompute this manual matrix. The linked audit worksheet records cell IDs, answer-artifact pointers, visible source URLs, and aggregate outcome flags, but it does not carry the complete provider, model, host, rule, and per-source role record. Attach companion protocol and source-role ledgers keyed by cell ID and report the result as a separate manual owned-link analysis.
Which inputs must be registered before classification?
Record the collection contract before reading the result so a convenient answer cannot silently change the denominator.
- Exact prompt text and prompt identifier, service, surface, locale, account or session state, exposed model label, and UTC collection time.
- Target host, URL normalization rule, and provider/surface-specific link-role rule, including redirects, canonical host aliases, subdomains, and which UI-marked citations or supporting sources qualify.
- Eligibility rule for observed cells and separate states for planned, unavailable, failed, and observed responses.
- A companion protocol ledger keyed to the worksheet cell ID, with one row for every planned cell: service or provider, surface, exposed model label or not available, target host, URL-normalization rule ID and text, and provider-and-surface link-role rule ID and text.
- A companion source-role ledger keyed to the same cell ID, with one row per visible URL: source order, UI label, UI role, UI location, original URL, resolved URL, normalized host, owned-host match, qualifying decision, exclusion reason, and capture pointer. Retain related or relevant links separately from qualifying citations.
- A declared distinction among a brand mention, recommendation, target-owned citation, other-source citation, and competitor replacement.
Which answer and citation states must stay separate?
Assign one availability state first, then record zero or more observed answer features. Do not turn collection failures into negative citation observations.
| State | Meaning | Denominator treatment |
|---|---|---|
| Planned | The registered cell has not been collected | Not observed; report separately |
| Unavailable | The service or declared surface could not return an eligible answer | Exclude from the observed denominator; retain the reason |
| Failed | Collection or parsing failed before an eligible answer was preserved | Exclude from the observed denominator; retain the error |
| Observed absence | An eligible answer was captured with no target mention or target-owned source | Include as an observed negative |
| Mention without target citation | The target is named, but no visible source resolves to the target-owned host | Include as an observed negative for target-owned citation |
| Target-owned citation | At least one qualifying UI-marked citation or supporting-source link resolves to the declared target-owned host | Include as an observed positive under that role and host rule |
| Target-owned related or relevant link | A visible target-owned link has a UI role that the frozen rule excludes from citations | Include in the observed denominator, exclude from the citation numerator, and retain the role |
| Other-source citation | A qualifying UI-marked citation or supporting-source link resolves elsewhere | Retain source type; do not relabel it target-owned |
| Competitor replacement | A declared competitor is named or cited where the target is absent | Retain as a separate observed feature |
Which synthetic cell ledger supports the worked matrix?
SYNTHETIC EXAMPLE — fictional brand, surfaces, sources, and observations; not RankEcho data, customer data, or a provider result.
Northstar S1 freezes three prompt jobs, two generic surfaces, and two repeats per prompt-surface pair. In this example, Surface A qualifies only inline links that its interface marks as citations; Surface B qualifies only links that its interface identifies as supporting sources. Merely related or relevant links are retained but excluded. “Common” means the same cell was observed in both windows; it does not mean the answer or citation state stayed unchanged.
| Cell and context | Baseline state and link role | Later state and link role | Common observed cell? |
|---|---|---|---|
| N01 · P1 category comparison · Surface A/R1 | Observed · qualifying target-owned citation or supporting source | Observed · qualifying target-owned citation or supporting source | Yes |
| N02 · P1 category comparison · Surface A/R2 | Observed · other source | Observed · other source | Yes |
| N03 · P1 category comparison · Surface B/R1 | Observed · no visible source | Observed · no visible source | Yes |
| N04 · P1 category comparison · Surface B/R2 | Unavailable | Unavailable | No |
| N05 · P2 calibration interval · Surface A/R1 | Observed · other source | Observed · qualifying target-owned citation or supporting source | Yes |
| N06 · P2 calibration interval · Surface A/R2 | Observed · no visible source | Observed · target-owned related link · excluded from citation numerator | Yes |
| N07 · P2 calibration interval · Surface B/R1 | Observed · no visible source | Observed · no visible source | Yes |
| N08 · P2 calibration interval · Surface B/R2 | Failed | Observed · qualifying target-owned citation or supporting source | No |
| N09 · P3 integration requirements · Surface A/R1 | Observed · qualifying target-owned citation or supporting source | Observed · qualifying target-owned citation or supporting source | Yes |
| N10 · P3 integration requirements · Surface A/R2 | Observed · mention only | Observed · mention only | Yes |
| N11 · P3 integration requirements · Surface B/R1 | Observed · other source · competitor replacement | Observed · other source · competitor replacement | Yes |
| N12 · P3 integration requirements · Surface B/R2 | Observed · no visible source | Observed · no visible source | Yes |
What does the worked synthetic gap matrix contain?
SYNTHETIC EXAMPLE — fictional brand, surfaces, sources, and observations; not RankEcho data, customer data, or a provider result.
Northstar S1 uses three prompts across Surface A and Surface B with two repeats each: 12 planned cells. The prompt rows below are computed from those registered cells before any cause is assigned.
| Prompt job | Planned | Unavailable | Failed | Observed | Owned-cited | Interpretation and next check |
|---|---|---|---|---|---|---|
| P1 · category comparison | 4 | 1 | 0 | 3 | 1 | Mixed; inspect comparison coverage without inferring cause |
| P2 · calibration interval | 4 | 0 | 1 | 3 | 0 | Complete observed gap; inspect claims to learn whether Fact Sourcing is warranted |
| P3 · integration requirements | 4 | 0 | 0 | 4 | 1 | Variable or weak; repeat and inspect prompt fit |
| Total | 12 | 1 | 1 | 10 | 2 | Observed cell-level owned-link rate is 2/10 = 20.0% |
How do you compare source coverage without inventing a cause?
Open each visible source and record which exact claims, entities, dates, definitions, and comparison fields it supports. Compare those fields with the target page. A missing supported field can justify a content inspection; it does not prove that adding the field changes selection.
Keep plausible alternatives visible: crawler or retrieval access, indexing eligibility, canonicalization, entity ambiguity, a prompt that does not fit the page, a stale fact, a different source type, or normal answer volatility. Select the smallest check that can distinguish among them.
Which false positives and collision risks should stop the diagnosis?
Stop if the record contains only a provider aggregate, a sitewide impression trend, an answer without preserved links, an unavailable or failed cell, or a brand mention that has been mislabeled as a citation. These observations answer different questions.
This page diagnoses a cell-level citation gap. Use the citation checker for a snapshot, citation monitoring for repeated product observations, competitor comparison for cross-brand patterns, and source intelligence for source inspection. Do not broaden this page into those jobs.
What deterministic validation completes the Find step?
The Find step is complete when the evidence record can be reproduced from the saved capture and yields one bounded next action. It does not wait for a later citation, referral, ranking, or traffic change.
| Check | Pass condition | Retained evidence |
|---|---|---|
| Cell identity | Prompt, service, surface, locale, account state, time, and matching rule are present | Registered cell record |
| Availability | Planned, unavailable, failed, and observed are never collapsed | State and reason |
| Source capture | Visible URLs, UI roles, and locations reproduce the classification | Raw answer, source panel, inline links, role labels, redirects, and normalized hosts |
| Semantic distinction | Mention, recommendation, target citation, other citation, and competitor replacement remain separate | Feature flags with definitions |
| Next check | One bounded hypothesis, its alternatives, and a result that distinguishes them are named without assigning cause | Reproducible check and decision rule; open a fix only for an established defect |
How should the gap feed the proof protocol?
Freeze the registered prompt-and-surface cells before a change. After the bounded fix, collect the same eligible cells and preserve unavailable, failed, unchanged, improved, and adverse results. Report matched cells separately from an available-case summary.
A later citation can be temporally associated with the change while other explanations remain possible. The raw numerator, denominator, exclusions, and cell intersection must remain visible.
Who maintains this analysis guide?
Written and maintained by Abiot Y. Derbie. Published 2026-09-07; sources checked 2026-09-06; last updated 2026-09-07. The worked matrix is synthetic and does not represent customer observations.
Recheck provider labels, source-display behavior, reporting definitions, and the visible evidence before applying the method. Send material corrections through the contact page.
Sources reviewed
Provider eligibility and measurement claims below were checked against primary documentation. These records do not establish a universal selection formula, causation, or a guaranteed ranking, impression, recommendation, or citation.
4 claim-level source records
| Claim reviewed | Official source | Review record |
|---|---|---|
| Bing's AI Performance report counts citations displayed as sources, reports cited-page activity, and exposes sampled grounding queries. Microsoft says those counts do not indicate placement, ranking, authority, page importance, or a page's role in an individual answer. | Bing Webmaster Blog: AI Performance | Checked 2026-09-06 · Public preview announced February 10, 2026 · Primary-source documentation check; Bing's provider-defined aggregate is kept separate from this guide's registered prompt-and-surface cells and from other providers' reporting units. · Confidence: High |
| Google's Generative AI performance report counts link impressions from AI Overviews and AI Mode and groups data by page, country, date, or device. Its documentation does not expose prompts, citation causes, a universal rank, or a source-selection formula. | Google Search Console: Generative AI performance report | Checked 2026-09-06 · Worldwide rollout stated as August 31, 2026 · Primary-source documentation check; a Google Search link impression is not relabeled as a fixed-cell citation, mention, recommendation, or matched prompt observation. · Confidence: High |
| Google says Gemini Apps may show inline links and a Sources button, but not every response has source links and a listed page can relate to only part of a response. | View related sources from Gemini Apps | Checked 2026-09-06 · Current Gemini Apps source guidance · Primary-source documentation check; a visible Gemini source is recorded only for the observed response and is not treated as a durable rank or claim-level support for the whole answer. · Confidence: High |
| OpenAI says ChatGPT responses that use web search may include citations and may expose cited sources and other relevant links in a Sources view. It also warns that search results and citations can be incomplete, outdated, or incorrect and should be checked against the opened source. | OpenAI Help: searching the web with ChatGPT | Checked 2026-09-06 · Current ChatGPT search source guidance · Primary-source documentation check; the observed visible citation is retained as evidence for that response, while source support and answer accuracy are checked separately. · Confidence: High |
Frequently asked questions
No. A mention names the target. A target-owned citation requires a qualifying UI-marked supporting or cited link that resolves to the declared target host under the recorded matching rule. Record recommendations separately too.
No. Report those collection states separately and exclude them from the observed citation denominator. Only an eligible observed answer can be a positive or negative citation observation.
No. It identifies an observed replacement and a source worth inspecting. It does not expose the system's private selection reason or prove that copying a field will produce a citation.
No. Traffic is a different, later observation. Classify a citation gap from a registered eligible answer and its preserved visible sources, without waiting for site traffic.
One reproducible cell-level classification, one bounded hypothesis or check, the plausible alternatives, and a result rule that determines whether the evidence supports a fix, another investigation, or no action.