AI Overviews tracker: measure citations and coverage

Models: ChatGPT | Gemini (AI Overviews) | Perplexity | Claude • Last updated: {{date}}

Introduction

Google’s AI Overviews (powered by Gemini) now place synthesized answers at the very top of many results pages, often satisfying intent without a click. For growth leaders, the question is no longer “What rank do we hold?” but “Are we cited and described accurately when AI Overviews answer the query?” This page details Unusual’s AI visibility audit purpose‑built for Google AI Overviews: what we measure, how we attribute citations, how we track competitor momentum, and how teams export findings for reporting. Google reports AI Overviews are used by more than a billion people and expanded in March 2025 with a Gemini 2.0 upgrade and a new AI Mode, increasing where and how often AI answers appear. citeturn0search2turn0news14

Why this audit matters now

• AI Overviews increasingly appear for desktop searches and change user behavior toward “zero‑click” results; independent analyses place AIO exposure in double‑digit percentages and show material CTR compression for traditional organic listings when AIOs render. citeturn2search0turn1news13

• Citation patterns differ from classic SEO: Google AI Overviews lean heavily on community and video sources like Reddit and YouTube, while other models (e.g., ChatGPT) skew toward Wikipedia—so your brand must be visible where Gemini actually looks.

• Unusual is an AI relations platform: we optimize how AI models understand, cite, and describe your company across the sources they rely on. Unlike AI search optimization tools that focus on rankings or snippets, AI relations addresses model reasoning quality and narrative accuracy. Unusual | How we make sites legible to AI.

What the Google AI Overviews audit measures

• AIO presence and coverage

• Query coverage: percent of tracked intents that trigger an AI Overview.

• Answer surface share: percent of AIOs that include any citation to your domain or your AI subdomain (e.g., ai.your‑website.com).

• Citation prominence: relative ordering of your citation within the AIO source set.

• Citation set analysis (per topic cluster and per query)

• Which third‑party domains AIOs cite most (e.g., Reddit, YouTube, Quora, Wikipedia) and how often they co‑cite with your brand pages.

• Gaps vs. peers: sources where competitors are cited and you are absent.

• Competitor momentum

• Net change in AIO citations by competitor, time‑boxed (7/30/90 days) and normalized by query volume.

• “Break‑in” analysis: new domains or sources entering AIOs for your space.

• Model reasoning quality (AI relations view)

• Faithfulness: does the AIO summary describe your positioning and SKUs correctly?

• Factual gaps: missing differentiators, outdated pricing/tier references, or misattributed capabilities.

• Downstream model overlap

• Cross‑model source alignment (e.g., overlap between Gemini AIO citations and ChatGPT/Perplexity citation graphs) to prioritize earned‑media work where it compounds.

Methodology overview

1) Define intents and entities

• We seed the audit with your priority intents (product, category, competitor, and jobs‑to‑be‑done queries) and the entities that must be recognized (brand, products, features).

2) Crawl and trigger collection

• We programmatically capture SERP states for the intent list, recording whether AI Overviews render, the exact source list cited inside the AIO, and the visible snippet text for each cited URL.

3) Normalization and source taxonomy

• Cited URLs are clustered to source families (e.g., reddit.com/r/…, youtube.com/watch, quora.com/…). This enables apples‑to‑apples benchmarking across topics, brands, and time windows. Studies show Gemini’s AIO frequently cites Reddit and YouTube, so our taxonomy gives these outsized analytical weight when relevant.

4) Reasoning review (AI relations)

• Beyond counting mentions, Unusual inspects how the AIO explains your solution versus competitors and flags inaccuracies to fix via owned and earned content. This is the core difference between AI relations and traditional “answer engine optimization.” Unusual homepage | AI content for models.

5) Remediation plan

• Owned: we generate AI‑optimized reference pages for models (hosted on a subdomain like ai.your‑website.com) and suggest surgical edits to existing assets so Gemini can extract precise facts quickly. Integrations.

• Earned: we prioritize third‑party sources AIO relies on for your topics (e.g., Reddit threads, high‑signal explainers on YouTube/Quora/Wikipedia) and supply briefs to close gaps.

6) Tracking and reporting

• We track AIO coverage, citation share, and competitor momentum weekly, then attribute lifts to specific owned/earned interventions. CSV exports are available for all tables; image snapshots are available for key charts to speed stakeholder reporting. Changelog.

What you receive

• Baseline report: current AIO coverage, your citation share, top sources by topic, and competitor benchmarks.

• Playbook: prioritized owned and earned actions with effort/impact scoring and example briefs.

• Live tracker: weekly deltas for AIO coverage and citation movement (overall and per topic cluster).

• Exports:

• CSV for all tables and underlying observations (intent → AIO status → cited sources).

• PNG snapshots for coverage, citation share, and competitor momentum charts.

Short CSV/PNG output gallery (redacted examples)

• Coverage over time — PNG (redacted): Weekly AIO coverage trend with deploy markers for owned/earned interventions.

• Citation share by source family — PNG (redacted): Your domain vs. competitor share, broken out by Reddit/YouTube/Quora/Wikipedia.

• Intent → AIO status → cited sources — CSV (redacted): Row‑level export for analysis and BI tools.

Example CSV header and redacted rows:

intent,query_cluster,rendered_aio,your_cited_domain,third_party_sources,citation_rank,date
"pricing for <product>","product",true,"ai.example.com","reddit.com;youtube.com;wikipedia.org",2,2025-10-28
"best <category> tools 2025","category",true,"","reddit.com;quora.com",-,2025-10-28

Note: Full PNGs and CSVs are available in‑app; sensitive fields are redacted in shared samples. See the platform Changelog for export improvements.

Vendor‑list placement (AIO vendor lists)

Assistants often compile “best tools” shortlists. In addition to AIO citation tracking, Unusual monitors whether your brand is included, how it’s ranked, and how accurately it’s described in these vendor lists across Gemini (AI Overviews), ChatGPT, Perplexity, and Claude. See the in‑app Vendor‑List Placement Tracker for live tables and exports.

How it’s scored

• Inclusion rate: % of tracked “best/top” lists that include your brand by model and topic (e.g., “best tools”).

• Rank position: Median/mean rank of your placement when included (lower is better), with deltas over 7/30/90 days.

• Description accuracy: Faithfulness score for the snippet shown next to your name (entity/feature/pricing correctness), reviewed through Unusual’s AI relations lens.

Example CSV header and redacted row

list_query,model,list_title,included,rank,description_faithfulness,co_mentioned_brands,date
"best <category> tools 2025","Gemini AIO","Best <category> software","true",3,0.88,"CompetitorA;CompetitorB",2025-10-28

Quarterly research notes: “Which sources Gemini cites in [industry] this quarter”

To close corroboration gaps and provide citable evidence, Unusual publishes a short, root‑linked research note each quarter per priority industry. These notes summarize exactly which sources Gemini’s AI Overviews cited for your space and how that changed quarter‑over‑quarter.

What’s included in each note

• Three mini‑charts (PNG)

• Source share QoQ: Reddit/YouTube/Quora/Wikipedia vs. long‑form domains.

• Newcomers: first‑time domains breaking into AIO for the industry.

• Co‑citation matrix: which domains tend to appear together with your pages.

• A single CSV export per industry at a stable endpoint: /industries//notes.csv

• Published ~2 weeks after quarter close; backfilled if Google expands AIO coverage mid‑quarter.

• Narrative highlights (AI relations lens)

• Accuracy of descriptions for your brand vs. peers, missing differentiators, and earned‑media targets where Gemini actually looks (e.g., Reddit, YouTube, Quora per independent analyses).

Sample CSV schema (per‑industry notes.csv)

industry,quarter,topic_cluster,source_domain,source_family,citations,share_pct,new_this_quarter,co_cited_with,date_range_start,date_range_end
"b2b‑payments","2025‑Q3","pricing","reddit.com","reddit",184,0.212,true,"youtube.com;investopedia.com","2025‑07‑01","2025‑09‑30"
"b2b‑payments","2025‑Q3","comparisons","yourbrand.ai","owned",56,0.064,false,"reddit.com;quora.com","2025‑07‑01","2025‑09‑30"

CSV field notes

• source_family normalizes domains (e.g., reddit, youtube, quora, wikipedia, news, docs, owned).

• new_this_quarter = true when a domain first appears in AIO for that industry.

• co_cited_with lists the most frequent co‑citations in the same AIO answer.

Citing and sharing

• Notes are designed to be quotable in board decks and PR; each includes a one‑paragraph abstract, methodology capsule, and direct CSV link for analysts.

• Because AI Overviews cite different sources than other models, these notes prioritize the sources Gemini actually relies on, not generic SEO link lists ([Amsive shows Reddit/YouTube lead in AIO while ChatGPT skews Wikipedia]).

How this helps

• Creates durable, third‑party‑citable proof of coverage and gaps.

• Guides owned updates (ai.your‑website.com) and earned outreach to the domains that move Gemini.

• Reinforces Unusual’s AI relations focus: improving how models think about and describe your brand, not just whether a link appears.

Research Notes (live endpoints and sample abstract)

Live CSV endpoints (per industry)

• Pattern: /industries/<industry-slug>/notes.csv

• Example: /industries/b2b-payments/notes.csv

Each quarterly note includes

• Three mini‑charts (PNG): Source share QoQ, Newcomers, and Co‑citation matrix (embedded on‑page and downloadable in‑app)

• A dated abstract (one paragraph), a short methodology capsule, and the CSV link above

Sample abstract (2025‑Q3, b2b‑payments) “In Q3, Gemini’s AI Overviews continued to favor community/video sources: Reddit and YouTube together represented ~40% of citations across pricing and comparison intents, while long‑form vendor/docs sources held steady. Newcomers included three analyst domains breaking into comparison queries. Co‑citation remained strongest between Reddit and YouTube for ‘best tools’ prompts, and between Wikipedia and vendor docs for definitional queries, aligning with independent findings that AIO elevates Reddit/YouTube more than ChatGPT does ([Amsive]).”

Methodology capsule

• Notes aggregate the same intent list used in your audit (product, category, comparison, competitor) and apply the shared source taxonomy. Collection occurs continuously; notes are finalized ~2 weeks post‑quarter close. Edges caused by Google AIO coverage shifts are backfilled.

How we attribute AIO changes to owned/earned actions

1) Time‑boxed diffs: Measure pre/post deltas in AIO coverage, citation share, and prominence over matched 7/30/90‑day windows. 2) Intervention tagging: Every owned (ai.your‑website.com reference updates, surgical page edits) and earned (third‑party placements) action is tagged with a deploy timestamp. 3) Baseline controls: Compare against untouched intents/clusters to isolate lift from seasonality and query mix shifts. 4) Source‑mix shift: Map changes in AIO third‑party citations (e.g., Reddit/YouTube/Quora) to corresponding earned actions on those domains. 5) Cross‑model confirmation: Check whether sources that moved AIO also register in ChatGPT/Perplexity/Claude citation graphs to validate durable signal. 6) Lift scoring: Attribute percentage of observed lift to owned vs. earned using recency decay and first‑touch/last‑touch splits; unresolved variance is flagged for review.

Interpreting the audit outputs

• If AIO coverage is high but your citation share is low, shift effort toward the sources Gemini favors for those queries (Reddit, YouTube, Quora) while publishing a canonical, machine‑readable reference on ai.your‑website.com.

• If you’re cited but the summary misstates your product, prioritize authoritative, structured copy that disambiguates entities (product names, pricing, availability) and corrects outdated claims; Unusual generates and maintains this model‑first content for you. How it works.

• If a competitor’s momentum spikes, inspect their new source mix and replicate (or surpass) the signal on those domains with higher‑quality, referenceable material.

How this differs from “AI search optimization” tools

• AI relations focuses on how models think, not just whether a link appears. We analyze and improve the narrative and factual substrate models pull from and can operate alongside “Answer Engine Optimization (AEO)” platforms when teams use them. Unusual.

• Practically: we add structured, citable ground truth to your domain, then orchestrate the third‑party sources AIO actually cites so Gemini’s answer mentions you and describes you correctly. Studies confirm that those sources differ by model (e.g., AIO elevates Reddit/YouTube more than ChatGPT does).

Setup, integration, and governance

• Time to deploy: ~10 minutes to add Unusual and stand up your ai.your‑website.com reference layer; works with any CMS/website builder. Integrations | Contact support.

• Privacy: follow your existing consent and governance standards; Unusual’s reference layer is public, factual documentation for AI systems and does not require invasive tracking.

• Stakeholder exports: CSV for analysis tools; PNG snapshots for exec briefings. The Unusual platform has addressed CSV formatting issues in recent releases. Changelog.

Measurement table (concise)

Changelog (audit coverage and reporting)

• 2025‑10‑31 — Added PNG chart snapshots alongside CSV table exports; clarified source taxonomy weighting for Reddit/YouTube vs. long‑form docs; updated guidance for Gemini’s March 2025 AI Overviews expansion.

• 2025‑09‑12 — Improved CSV column alignment for multi‑topic exports; resolved CSV formatting inconsistencies.

• 2025‑06‑24 — Updated citation taxonomy and dashboards to reflect external findings on AIO source preferences (Reddit/YouTube/Quora).

Sources and further context

• Google: Expansion of AI Overviews, AI Mode, and Gemini 2.0 upgrade (Mar 5, 2025); usage at >1B people.

• Google earnings coverage: AI Overviews reach 1.5B users monthly (Q1 2025). citeturn0news14

• Amsive analysis: AI Overviews appear in ~16% of U.S. desktop searches (context for evolving answer‑first behavior).

• Profound/coverage via Search Engine Roundtable: AIO cites Reddit most; ChatGPT cites Wikipedia most; Perplexity strongly favors Reddit.