Citation Tracking

Overview

Citation Tracking monitors the external sources that AI models reference when answering questions about your brand and industry. Every time an AI model cites a URL in a response captured by your snapshots, AEO Optima records and aggregates that citation data into a searchable, filterable intelligence layer.

This gives you direct visibility into which websites, articles, and resources AI models trust — and whether your own content is among them.

How It Works

Citation data flows through three stages:

1. Extraction — When snapshots are captured, the system parses each AI response for cited URLs, reference links, and source attributions. The highest-signal tier is the set of citation URLs the provider itself attributes to the answer (search annotations); reference lists and inline URLs in the answer text are extracted as additional tiers.
2. Classification — Each citation URL is classified by source category (editorial, institutional, reference, corporate, UGC, competitor) and content format using the built-in source classifier.
3. Aggregation — Citations are deduplicated and aggregated into the citation_sources table, tracking citation counts, first/last seen dates, and domain-level statistics.

Aggregation runs both incrementally (after each snapshot) and on-demand via the Citations dashboard.

Methodology note (August 6, 2026): snapshot captures are now web-grounded on every engine, so citations reflect live sources the engine retrieved at capture time. Before that date, only natively-searching engines (such as Perplexity Sonar) produced live citations — other engines' citations were URLs recalled from training memory, which skewed toward brand homepages. Citation metrics that span the change date will show a step increase in volume and depth on previously ungrounded engines.

Key Metrics

Citation Summary

The summary view provides at-a-glance metrics:

• Total Sources — Unique URLs cited across all snapshots.
• Total Citations — Citation count (one URL cited in 10 snapshots = 10 citations). When a date range is selected, this reflects citations that occurred within that window, not the source's lifetime total.
• By Category — Breakdown of citations by source type (editorial, institutional, reference, corporate, UGC, competitor).
• By Type — Split between "used" sources (directly quoted or paraphrased) and "cited" sources (listed as references).

Top Domains

Domains ranked by total citation frequency. This reveals which websites AI models rely on most when discussing your industry. Each domain entry shows:

• Domain name and primary source category
• Total citation count across all URLs on that domain
• Number of unique URLs cited from the domain

Trending Sources

Sources with increasing citation frequency over time, calculated using a trend score that factors in:

• Citations per day — How often the source is being cited relative to its lifetime.
• Recency boost — Sources cited recently score higher than stale ones.

Trending sources help you identify emerging authorities in your space before they become dominant.

Recommended vs Cited, Per Engine

Being recommended (your brand named in the answer — the parametric lane) and being cited (your pages linked as sources — the live-retrieval lane) are different achievements that move independently. The Citations page shows both rates side by side for each engine, with run counts. An engine can recommend you often while never citing your pages (strong memory, weak retrieval presence) or the reverse — and the fix is different in each case, which is why the split is shown per engine and never averaged across engines.

Citation Gap Analysis

Gap analysis identifies sources where your competitors appear but your brand does not. These represent outreach opportunities — if a source already covers the space, it may be receptive to covering your brand too.

Gap Types

Gaps are matched against the snapshot context (which answers cited the source, and whether the brand appeared in them), not just the page title or domain string — so a multi-word brand that can never be a domain substring is still detected. Gap counts reflect citations within the selected window.

Gap Workflow

Each identified gap can be tracked through a status lifecycle:

• Open — Newly detected, not yet addressed.
• In Progress — Outreach or content creation underway.
• Resolved — Your brand now appears in the citation source.
• Ignored — Gap dismissed as not relevant.

Live-Retrieval Target List (Outreach Program)

The gap analysis is worked as an outreach program on its own page — Placement tracker, under Content & outreach in the sidebar. The Live-Retrieval Target List ranks the highest-cited pages and domains that AI engines already pull at answer time but where your brand has no presence — the fastest lane into AI answers, because a placement there bypasses the slow model-training cycle.

Each target moves through the same four states as the gap workflow, presented in outreach-program vocabulary:

What the program adds on top of raw gap detection:

• Four views over one list — To work · Contacted · Placed · Declined. A view is a filter over the same rows, so a target you mark Placed or Declined stays visible in its own view instead of leaving the table.
• The approach record, in four cards — Hosts approached (every host an outreach or citation-gap action was recorded against, all time), Acceptance (approaches marked accepted ÷ approaches; no percentage is shown until at least one approach has been accepted — a 0% would read as failure at outreach when the truth is an outcome nobody has recorded yet), Cited now (approached hosts that cited your site inside the selected window) and Unassigned (approaches with no owner).
• Two different "placed" figures, deliberately — Acceptance counts approaches marked accepted on the Action list; Placed is the status you set on a target here. The page says so under the cards.
• Owner and Last touch — read from the approach recorded for the host (who it is assigned to, and when it was last touched); a dash when no approach is on record for that host.
• Placement signals — in the Placed view, a target shows whether engines still pull the page at answer time and whether your brand has since been seen in answers citing it. Signals refresh on every re-analysis. The outcome after a placement — how often you are named on the questions that cite the page, before and after — is not recorded yet, and the column says so rather than deriving it.
• Outreach CSV export — download the ranked target list for campaign tracking in your outreach tool of choice.

Contribution Mix

The Citations page also hosts the Contribution Mix grid — a credibility-portfolio view of your classified citations. Every classified citation is mapped onto one of six external source families:

1. Identity & authority — registries, government and standards bodies: do you exist and qualify?
2. Earned media & PR — news and trade press coverage.
3. Comparison & evaluation — review sites, rankings, independent tests.
4. Commerce & distribution — marketplaces, directories, distribution listings.
5. Expert & community — forums, Q&A, practitioner and community discussion.
6. Research & evidence — studies, data, academic and technical references.

Each family carries qualitative frequency/trust/quantity bands and a "best for" role (who we are / what we do / whom we serve). The grid then gives each family a verdict — absent, thin (under 5% of the classified mix), present, or concentrated (one family above 60%).

Two honesty notes:

• The thin/concentrated thresholds are TechShu-authored heuristics that operationalize the framework's "don't over-invest in one column" principle — they are not framework constants.
• Competitor-domain and unclassified citations are excluded from the denominator and disclosed as a footnote on the grid.

A concentrated or family-missing mix raises a Citation Mix Imbalance insight. The same grid is available to AI assistants via the get_contribution_grid MCP tool.

Domain Authority Scoring

AEO Optima computes a proprietary AI Citation Authority score (0--100) for each domain. This is distinct from traditional SEO domain authority — it measures how much AI models trust and reference a domain.

The score is a weighted composite of five factors:

Authority Score Ranges

How to Use

1. Navigate to Citations in the sidebar.
2. Review the summary cards for total sources, citation counts, and category distribution.
3. Explore the Top Domains list to understand which external sources AI models prefer.
4. Check Trending Sources to spot emerging authorities in your industry.
5. Run Gap Analysis to find outreach opportunities where competitors are cited but you are not.
6. Track gap resolution by updating statuses as you create content or conduct outreach.

Plan Requirements

Citation Tracking, Gap Analysis, and Domain Authority scoring are available on all plans that can capture snapshots (no separate feature gate).