# PeakMetrics

> PeakMetrics is a narrative intelligence platform that shows communications, public affairs and risk teams how stories about their organization spread across news, social media and AI answers, and flags when those narratives are coordinated or manipulated. Its AI Perceptions product measures how ChatGPT, Claude, Gemini, Perplexity and Grok describe a brand, which is the practice usually called Generative Engine Optimization (GEO). The company's position is that press coverage and AI visibility are different things, and that comms teams need to measure both.

This is the long version of https://www.peakmetrics.com/llms.txt. Everything below is either a description of what PeakMetrics offers or a dated measurement with its limits stated next to it.

## What PeakMetrics does

PeakMetrics monitors news and social media coverage and shows how narratives form and spread. Customers are communications, public-affairs, risk and government teams. The platform is described at https://www.peakmetrics.com/platform, and the government use case at https://www.peakmetrics.com/use-cases/gov.

### AI Perceptions

AI Perceptions is the product for GEO. Where a rank tracker reports how often a brand appears, AI Perceptions is built to show why an assistant says what it says and which sources it draws on, alongside the news and social narratives PeakMetrics already tracks. Details at https://www.peakmetrics.com/ai-perceptions.

- Benchmark questions are run daily across five providers: ChatGPT, Claude, Gemini, Perplexity and Grok.
- Answers are analyzed for how the brand is described (Favorability) and sorted with Smart Categories, which customers can define themselves, for example purchase intent or brand alignment.
- Cited domains and pages show which sources the answers draw on.

### Other products and pages

- Platform: https://www.peakmetrics.com/platform
- Intelligence reports: https://www.peakmetrics.com/solutions/intelligence-reports
- Government: https://www.peakmetrics.com/use-cases/gov
- Public affairs: https://www.peakmetrics.com/solutions/public-affairs (answers: How can a public affairs team track how a policy issue is spreading across news and social?)
- Coordinated attack detection: https://www.peakmetrics.com/solutions/coordinated-attack-detection (answers: How do I tell if a wave of negative posts about my company is coordinated, and prove it?)
- Case studies: https://www.peakmetrics.com/case-studies
- Insights (articles): https://www.peakmetrics.com/insights/page/1
- Responsible AI: https://www.peakmetrics.com/responsible-ai
- Demo: https://www.peakmetrics.com/demo
- Contact: https://www.peakmetrics.com/contact

## Free tools

- [Fact Check topics](https://www.peakmetrics.com/free-tools/fact-checks#topics): Browse published reviews by subject without signing in. Original ratings belong to the cited publishers; PeakMetrics summaries and context are AI-assisted and may be pending or unavailable. Cite the original review and its date, and distinguish it from PeakMetrics research and retrieval dates. The public corpus sitemap is https://www.peakmetrics.com/sitemap.fact-checks.xml.

- [Fact Checks](https://www.peakmetrics.com/free-tools/fact-checks): Search published fact checks by text or image and read attributed publisher ratings, cited summaries, and research. Website search needs no account. REST API and MCP access require a verified email and API key. Integration documentation: https://www.peakmetrics.com/free-tools/fact-checks/docs.

- https://www.peakmetrics.com/free-tools: index.
- https://www.peakmetrics.com/free-tools/ai-perceptions: the free AI Perceptions check, described below.
- https://www.peakmetrics.com/free-tools/ai-perceptions/citation-sources: which domains AI assistants and Google AI Overviews cite for corporate reputation and consumer shopping questions.
- https://www.peakmetrics.com/free-tools/daily-brief-generator: enter a brand, get a summary of its news coverage from the last 72 hours, with each claim linked to the source article. Limited to 10 briefs a day per browser.
- https://www.peakmetrics.com/glossary: definitions of GEO, AI visibility, cited domains and related terms.

## Guides

- https://www.peakmetrics.com/guides/media-monitoring-tools: buyer's guide to media monitoring tools.
- https://www.peakmetrics.com/guides/social-listening-tools: buyer's guide to social listening tools.
- https://www.peakmetrics.com/guides/correct-ai-answers: playbook for correcting what AI assistants say about an organization: sample answers, log errors, trace and fix the sources, use provider feedback channels, re-check on a schedule.
- https://www.peakmetrics.com/guides/geo-for-comms-teams: playbook on who owns GEO across comms, SEO, marketing, legal, IR and product, with a 90-day plan and what to measure.

## The free AI Perceptions check: what it measures and what it does not

The free check publishes pre-computed snapshots for 20 sectors and 298 brands. Each sector has 250 unbranded questions spread across intents such as reliability, value and customer service (for example "Which US airlines are the most reliable?"). A model drafted the questions, a review pass removed paraphrases and questions that point to one company, and every question was checked against the name of every measured brand. Each question was put once to each of ChatGPT, Claude, Gemini, Perplexity and Grok, giving 1,250 answers per sector and 25,000 in all. The brand is never named in the question, so an appearance means the assistant volunteered it.

Limits that apply to every figure below:

- Latest run: 2026Q4, 2026-09-29. The benchmark is refreshed quarterly, not daily.
- It asks the same five assistants the paid product benchmarks, on their product models, with web search on. Each question is asked once per assistant.
- Every visibility figure has a 95% range (a Wilson interval). Where two brands' ranges overlap, treat them as tied.
- These are snapshots, not rankings or trends. Live model answers vary between runs.
- In the AI companies sector, an assistant naming its own maker is not counted (Claude's answers for Anthropic, ChatGPT's for OpenAI and so on); those brands' figures rest on the other assistants' answers.
- The quotes in each sector story come from a stratified sample of the answers, spread across every assistant and intent. The figures use every answer.
- Visibility is the share of a sector's answers that named a brand. It is not a favorability score and says nothing about how the brand was described.
- Source-type and topic labels on these pages come from a classifier run on the PeakMetrics marketing site. They are not a customer workspace's Smart Categories.

### Sector snapshots

- [Aerospace & Defense](https://www.peakmetrics.com/free-tools/ai-perceptions/aerospace-defense), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Boeing 53% (95% range 51-56%, 667 of 1,250 answers); Lockheed Martin 42% (95% range 40-45%, 529 of 1,250 answers); Northrop Grumman 35% (95% range 32-37%, 435 of 1,250 answers).
- [AI Companies](https://www.peakmetrics.com/free-tools/ai-perceptions/ai-companies), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: OpenAI 69% (95% range 66-72%, 694 of 1,000 answers); Google DeepMind 69% (95% range 66-72%, 691 of 1,000 answers); Anthropic 66% (95% range 63-69%, 662 of 1,000 answers).
- [Airlines](https://www.peakmetrics.com/free-tools/ai-perceptions/airlines), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Delta Air Lines 67% (95% range 65-70%, 843 of 1,250 answers); United Airlines 60% (95% range 57-63%, 748 of 1,250 answers); Southwest Airlines 46% (95% range 43-49%, 575 of 1,250 answers).
- [Automotive](https://www.peakmetrics.com/free-tools/ai-perceptions/automotive), measured 2026-09-29, 16 brands, 250 questions, 1,250 answers. Most visible: Toyota 56% (95% range 53-58%, 697 of 1,250 answers); Ford 45% (95% range 42-48%, 560 of 1,250 answers); Tesla 45% (95% range 42-47%, 557 of 1,250 answers).
- [Beauty & Personal Care](https://www.peakmetrics.com/free-tools/ai-perceptions/beauty-personal-care), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Sephora 26% (95% range 23-28%, 321 of 1,250 answers); Ulta Beauty 24% (95% range 22-27%, 301 of 1,250 answers); L'Oréal 24% (95% range 22-26%, 300 of 1,250 answers).
- [Consumer Packaged Goods](https://www.peakmetrics.com/free-tools/ai-perceptions/consumer-packaged-goods), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Procter & Gamble 25% (95% range 23-28%, 314 of 1,250 answers); Unilever 22% (95% range 19-24%, 269 of 1,250 answers); Nestle 21% (95% range 19-24%, 266 of 1,250 answers).
- [Cybersecurity](https://www.peakmetrics.com/free-tools/ai-perceptions/cybersecurity), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Microsoft Security 42% (95% range 39-44%, 520 of 1,250 answers); CrowdStrike 35% (95% range 33-38%, 442 of 1,250 answers); Palo Alto Networks 32% (95% range 30-35%, 404 of 1,250 answers).
- [Energy & Utilities](https://www.peakmetrics.com/free-tools/ai-perceptions/energy), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: ExxonMobil 14% (95% range 12-16%, 169 of 1,250 answers); Shell 13% (95% range 11-15%, 160 of 1,250 answers); Chevron 13% (95% range 11-15%, 158 of 1,250 answers).
- [Financial Services](https://www.peakmetrics.com/free-tools/ai-perceptions/financial-services), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: JPMorgan Chase 47% (95% range 45-50%, 591 of 1,250 answers); Bank of America 38% (95% range 35-40%, 469 of 1,250 answers); Wells Fargo 30% (95% range 27-32%, 372 of 1,250 answers).
- [Fintech & Payments](https://www.peakmetrics.com/free-tools/ai-perceptions/fintech-payments), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: PayPal 43% (95% range 40-45%, 532 of 1,250 answers); Block 41% (95% range 38-43%, 508 of 1,250 answers); Stripe 22% (95% range 19-24%, 269 of 1,250 answers).
- [Gaming](https://www.peakmetrics.com/free-tools/ai-perceptions/gaming), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Sony PlayStation 47% (95% range 44-49%, 583 of 1,250 answers); Xbox 41% (95% range 39-44%, 515 of 1,250 answers); Nintendo 40% (95% range 38-43%, 503 of 1,250 answers).
- [Healthcare & Pharma](https://www.peakmetrics.com/free-tools/ai-perceptions/healthcare-pharma), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Johnson & Johnson 15% (95% range 13-17%, 192 of 1,250 answers); UnitedHealth Group 14% (95% range 12-16%, 179 of 1,250 answers); CVS Health 14% (95% range 12-16%, 174 of 1,250 answers).
- [Hotels & Travel](https://www.peakmetrics.com/free-tools/ai-perceptions/hotels-travel), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Marriott 55% (95% range 52-58%, 687 of 1,250 answers); Hilton 54% (95% range 52-57%, 681 of 1,250 answers); Hyatt 43% (95% range 40-46%, 539 of 1,250 answers).
- [Insurance](https://www.peakmetrics.com/free-tools/ai-perceptions/insurance), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: State Farm 39% (95% range 36-41%, 483 of 1,250 answers); Allstate 29% (95% range 27-32%, 367 of 1,250 answers); Progressive 28% (95% range 26-30%, 349 of 1,250 answers).
- [Logistics & Delivery](https://www.peakmetrics.com/free-tools/ai-perceptions/logistics), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: FedEx 72% (95% range 69-74%, 898 of 1,250 answers); UPS 71% (95% range 68-73%, 885 of 1,250 answers); USPS 47% (95% range 44-50%, 589 of 1,250 answers).
- [Restaurants & Fast Food](https://www.peakmetrics.com/free-tools/ai-perceptions/restaurants), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: McDonald's 51% (95% range 48-53%, 634 of 1,250 answers); Chick-fil-A 38% (95% range 35-40%, 472 of 1,250 answers); Chipotle 35% (95% range 32-38%, 437 of 1,250 answers).
- [Retail](https://www.peakmetrics.com/free-tools/ai-perceptions/retail-hospitality), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Walmart 46% (95% range 43-49%, 577 of 1,250 answers); Target 36% (95% range 34-39%, 452 of 1,250 answers); Costco 34% (95% range 31-37%, 423 of 1,250 answers).
- [Ride-Hailing & Delivery](https://www.peakmetrics.com/free-tools/ai-perceptions/ride-hailing-delivery), measured 2026-09-29, 12 brands, 250 questions, 1,250 answers. Most visible: Uber 86% (95% range 83-87%, 1,069 of 1,250 answers); Lyft 51% (95% range 48-54%, 636 of 1,250 answers); DoorDash 42% (95% range 40-45%, 530 of 1,250 answers).
- [Technology](https://www.peakmetrics.com/free-tools/ai-perceptions/technology), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Google 65% (95% range 62-68%, 813 of 1,250 answers); Apple 45% (95% range 43-48%, 566 of 1,250 answers); Microsoft 42% (95% range 40-45%, 528 of 1,250 answers).
- [Telecom & Media](https://www.peakmetrics.com/free-tools/ai-perceptions/telecom-media), measured 2026-09-29, 15 brands, 250 questions, 1,250 answers. Most visible: Verizon 48% (95% range 45-51%, 599 of 1,250 answers); AT&T 46% (95% range 43-49%, 576 of 1,250 answers); T-Mobile 40% (95% range 38-43%, 506 of 1,250 answers).

## Findings on cited sources

Scope: 25200 answers from Perplexity, Gemini, ChatGPT, Claude, Grok (25000 answers to 5000 corporate reputation questions across 20 sectors, and 200 answers to 40 consumer shopping questions), containing 151559 citations to 34802 unique domains, measured 2026-09-30. Each assistant was called through its API with web search on, on the model the paid product benchmarks. The consumer apps can use different models and search settings. Google AI Overviews for the same questions (93 of 105 questions triggered one; DataForSEO SERP google/organic/live/advanced, US English, desktop) are reported separately and never pooled with the assistants. One snapshot, not a trend.

- Reddit on two surfaces: on shopping questions Reddit was 7.8% of citations in Google AI Overviews (in 40.6% of overviews) and 1.1% of citations across the five assistants (in 7.5% of answers). On corporate questions it was 3.9% in AI Overviews and 0.9% across the assistants. The claim that Reddit dominates AI answers holds for AI Overviews on shopping questions and mostly does not for the assistants.
- The assistants differ from each other: Perplexity returned 8.8 cited domains per answer and cited Reddit in 6.6% of its answers; Gemini returned 6.9 cited domains per answer and cited Reddit in 16.7% of its answers; ChatGPT returned 2.6 cited domains per answer and cited Reddit in 0.4% of its answers; Claude returned 3.5 cited domains per answer and cited Reddit in 0% of its answers; Grok returned 8.4 cited domains per answer and cited Reddit in 5% of its answers. Perplexity returns far more cited domains per answer than ChatGPT, so pooled shares lean toward the heaviest citers. Provider-balanced Reddit share (mean of each assistant's own share) was 0.8% on shopping questions and 0.8% on corporate questions.
- Most-cited domains across the assistants, pooled: forbes.com (1706 citations, 1.1%); reddit.com (1441 citations, 1%); consumerreports.org (925 citations, 0.6%); nerdwallet.com (787 citations, 0.5%); youtube.com (780 citations, 0.5%); businessinsider.com (759 citations, 0.5%); en.wikipedia.org (647 citations, 0.4%); cnbc.com (621 citations, 0.4%). No domain accounts for more than 1.1% of citations, so the citations are spread across a long tail. On shopping questions the leaders were forbes.com (2.5%), nerdwallet.com (2%), nytimes.com (1.7%), consumerreports.org (1.6%), pcmag.com (1.5%); on corporate questions forbes.com (1.1%), reddit.com (0.9%), consumerreports.org (0.6%), youtube.com (0.5%), nerdwallet.com (0.5%).
- Google AI Overviews lean on social and user-generated platforms. Most-cited domains on shopping questions: reddit.com (7.8%), youtube.com (6.6%), facebook.com (3.6%), nerdwallet.com (3%), quora.com (2.4%). On corporate questions: facebook.com (6%), youtube.com (5%), reddit.com (3.9%), quora.com (2.8%), instagram.com (2.1%).
- Wire services: 475 assistant citations (0.3% of the pool) were direct links to wire-service domains (PR Newswire, Business Wire, GlobeNewswire and similar). This is a lower bound, not a total. Of 125896 cited pages checked for syndicated wire copy, 96386 could be read and 79 were confirmed as syndicated wire releases; 29510 could not be read because the sites blocked automated requests. Those 29510 are unknown, not confirmed absent.
- Community sites (Reddit, Quora, Stack Exchange, Hacker News, Trustpilot, Glassdoor): 2564 assistant citations (1.7%), of which 1441 (1%) were from reddit.com.

Full detail and method: https://www.peakmetrics.com/free-tools/ai-perceptions/citation-sources.

## How PeakMetrics describes itself in this file

Descriptions of paid products above are summaries. For current capabilities, pricing and availability, see https://www.peakmetrics.com/ai-perceptions and https://www.peakmetrics.com/demo, or contact https://www.peakmetrics.com/contact.