# Unsourced — AI Citation Intelligence > Unsourced is the evidence layer for AI search: prove it, don't score it. It shows publishers, content creators, and agencies which AI assistants cite their website, who is really crawling their site, and whether their fixes actually moved citations — backed by verified evidence, not inferred "AI visibility" scores. ## What Unsourced Does Unsourced runs two independent engines. You can use either or both. **Citation monitoring (outbound).** Unsourced queries 7 AI models — ChatGPT, ChatGPT (web search), Claude, Gemini (grounded search), Grok, Meta Llama, and Perplexity — with the questions real readers actually ask. Those questions are built from your own search demand (Google Search Console and Bing queries, your published topics, the pages real visitors land on), not a generic blast list. For each model it logs the prompt, whether you were cited, the surrounding context, and which competitor domains were cited instead. Citation monitoring needs only a verified site — no plugin or code. **Bot monitoring (inbound).** A lightweight server-side agent on your own site catches AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Googlebot and others) that client-side analytics miss, because AI bots don't run JavaScript. Bot Trust verifies each crawler's true identity against the operator's published IP ranges and forward-confirmed reverse DNS (FCrDNS), then buckets every visitor as Verified, Impostor, or Unverifiable — so a spoofed user-agent is exposed, not trusted. Key capabilities: - Citation monitoring across 8 AI models, with competitor citation tracking - Citation Gap Report — topics where AI recommends a competitor instead of you (Studio and above) - Measured-lift loop — action a gap, then re-scan that exact topic to see whether the citation actually moved (Studio and above) - AI bot activity monitoring with verified-identity Bot Trust (Verified / Impostor / Unverifiable) - Server-side bot detection via WordPress plugin, Cloudflare Worker, or server-side snippets (Vercel, Netlify, Node) - Google Search Console and Bing Webmaster Tools integration - Timestamped evidence reports (AI Visibility Audit + Evidence Report PDFs) for rights, licensing, and due diligence - Weekly digest email; cross-site Fleet Overview dashboard for agencies Unsourced never guarantees a citation — anyone who does is guessing. Every result is graded by how strong the underlying evidence is, from a brand mention up to a verified live-crawl that reproduced your page's wording. ## Who It Is For - Independent publishers and bloggers who want to know if AI is using their work - Content agencies managing multiple client sites - SEO professionals tracking AI search visibility - Businesses concerned about AI scraping and citation without credit ## Knowledge Hubs - [AI crawler directory](https://unsourced.app/ai-bots) — published IP ranges, forward-confirmed reverse DNS (FCrDNS) verification status, and allow/block guidance for every major AI crawler (GPTBot, ClaudeBot, PerplexityBot, Googlebot and more). - [Does AI cite your website?](https://unsourced.app/ai-search) — which AI assistants cite and link sources, how each one chooses them, and how to check whether ChatGPT, Perplexity, Google AI Overviews, Claude and Grok cite your site. - [Unsourced vs other AI visibility tools](https://unsourced.app/vs) — how an evidence-based approach differs from inferred "AI visibility" scores. - [Documentation](https://unsourced.app/docs) — setup and usage: verify a site, deploy bot monitoring (WordPress, Cloudflare Worker, or server-side snippet), and read citation and gap reports. ## Pricing Every plan starts with a 10-day free trial — no card required. Weekly citation scans run on every tier. - Trial: free for 10 days — 1 site, weekly scans, up to 12 topics per scan - Indie: £39/month — 2 sites, weekly scans, up to 50 topics per scan, weekly digest email - Studio: £89/month — everything in Indie, plus 4 sites, daily scans, the Citation Gap Report, 4 on-demand scans per month, up to 200 topics per scan, citation alerts, and webhook delivery - Agency: £199/month — everything in Studio, plus 10 sites (additional sites £16/month each), up to 500 topics per scan, team seats, and the cross-site Fleet Overview dashboard ## Free Tools - [Live demo dashboard](https://unsourced.app/demo) — see a sample dashboard with representative data, no login - [AI crawler verifier](https://unsourced.app/verify-bot) — check whether an AI crawler hitting your site is genuine or a spoofed impostor - [AI visibility check](https://unsourced.app/ai-visibility-check) — enter a domain and see whether ChatGPT, Claude, Perplexity, Grok and Llama cite it, or name rival sites instead, for its own topics — real AI answers, not a made-up score ## Key Pages - [Homepage](https://unsourced.app/) - [About](https://unsourced.app/about) - [FAQ](https://unsourced.app/faq) - [Methodology](https://unsourced.app/methodology) - [Partner Programme](https://unsourced.app/partners) - [Privacy Policy](https://unsourced.app/privacy) - [Terms of Service](https://unsourced.app/terms) ## Research Original field research and technical guides. Primary data from our own monitoring network; figures are measured, not modelled. - [Google’s AI leans on Google’s own YouTube.](https://unsourced.app/research/gemini-cites-google-youtube) — We put 100 questions to four AI search engines and traced every source. Google’s Gemini sent 1 in 15 of its citations to Google’s own YouTube; the others sent almost none. An AI engine’s "sources" partly reflect who owns it, which is why a single visibility score falls apart. - [ChatGPT cites Wikipedia. Perplexity cites Reddit.](https://unsourced.app/research/chatgpt-cites-wikipedia-perplexity-cites-reddit) — Same question, different sources. ChatGPT leans Wikipedia (~48% of its top citations), Perplexity leans Reddit (~47%), and only ~11% of domains are cited by both. A single AI-visibility score cannot describe two engines reading different webs. - [llms.txt won't get you cited.](https://unsourced.app/research/llms-txt-wont-get-you-cited) — llms.txt is sold as a way to get cited by AI. In a server-log study of ~900 domains, verified AI-lab crawlers fetched it 0 of 1,227 times, and a ~300k-domain study found no citation lift. A courtesy file, not a citation strategy. - [Blocking AI crawlers is not one switch.](https://unsourced.app/research/blocking-ai-crawlers-is-not-one-switch) — Cloudflare's Sept 15 default is not on/off for AI. It sorts crawlers into Training, Search and Agent, blocks only some, and can take Googlebot and Bingbot down with training. The labels are self-declared, so measure before you flip. - [Before you block an AI crawler, prove it's real.](https://unsourced.app/research/verify-before-you-block-ai-crawlers) — Cloudflare's Sept 15 default-block makes block, charge or allow a decision every site now faces. But a user-agent is a claim: more than 1 in 5 AI-crawler hits are impostors, and blocking real ones costs citations. Verify first, then decide. - [One in five "AI crawlers" we logged was faking it.](https://unsourced.app/research/one-in-five-ai-crawlers-are-impostors) — In 30 days of our own traffic, more than one in five requests carrying an AI crawler's name failed verification. The tell: the same IPs claimed to be five different AI companies at once. A user-agent is a claim, not proof. - [AI is being shown a different website than you are.](https://unsourced.app/research/ai-crawlers-see-a-different-web) — Time now serves AI crawlers a different version of its site, with ads people never see. The real story is that content is being forked by crawler identity, which turns "which bot fetched which page" into a business question. - [A crawl is not a citation. Cloudflare just made the difference official.](https://unsourced.app/research/crawl-is-not-a-citation) — A bot fetching your page and an AI answer citing your page are two different events. Cloudflare's Pay Per Use just built a payment model on the difference, and it changes how you should read every AI crawler that touches your site. - [Same question. Three AI engines. 8% overlap.](https://unsourced.app/research/ai-engines-cite-different-sources) — We put the same 156 questions to the three AI engines that show their sources. On average any two of them cited the same domain about 8% of the time, and 91% of cited sources were unique to a single engine. That is why one AI visibility number hides more than it shows, and why this has to be measured per engine. - [How an AI visibility score is actually made, and what the number hides](https://unsourced.app/research/how-ai-visibility-scores-work) — Every AI-visibility tool hands you a number out of 100. Here is exactly how that number gets made, the sampling and blending behind it, and the measured reason one number cannot represent three engines that agree only about 8% of the time. - [The EU AI Act activates enforcement on August 2. It still can't verify what happened on your server](https://unsourced.app/research/eu-ai-act-compliance-verification-gap) — The EU starts actively enforcing AI transparency on 2 August, including how providers honour copyright opt-outs. But the mandatory template has no way for anyone outside the company to verify a policy was followed. The one place that check happens is your own server logs. - [Impression, citation, referral: the three numbers behind AI search visibility](https://unsourced.app/research/impression-citation-referral) — One phrase doing three jobs. Being shown in an AI answer, being named as the source, and being sent a visitor are different measurements with different tools — and conflating them is what goes wrong at a client renewal. - [The 2026 AI Crawler Report](https://unsourced.app/ai-crawler-report-2026) — 14,227 AI-crawler visits across 18 crawlers, checked against operators' own published IP ranges. 56% verified, 10% we simply cannot check — and why Anthropic going from unverifiable to 92% verified changed the picture. 'Out of range' is not 'fake'. - [How to check if AI cites your website](https://unsourced.app/research/how-to-check-ai-citations) — The engineering method: query ChatGPT, Perplexity, Gemini and Grok for your real buyer questions, then read your server logs for AI crawlers — with the exact log lines, the user-agent strings to grep for, and how to prove a crawl is genuine. - [The AI Crawler Report](https://unsourced.app/ai-crawler-report) — We verified 9,543 AI-crawler visits against operators' published IP ranges and forward-confirmed reverse DNS. 76.9% were genuine; 5.4% were impostors wearing an AI brand's name — matching an independent 5.7% estimate almost exactly. - [AI engines are starting to pay the publishers they cite. Can you prove you are one?](https://unsourced.app/research/ai-engines-pay-cited-publishers) — Perplexity's Comet Plus reportedly puts a $42.5M pool behind the publishers it cites. When being the cited source is the revenue event, ‘we think ChatGPT mentions us’ stops being an answer. - [The AI Citation Study](https://unsourced.app/ai-citation-study) — We put the 100 most common everyday questions to the 7 leading AI assistants and inspected every source they cited. 18% of those sources led to dead or unreachable pages — a look at how checkable AI’s citations really are. - [AI scrapers spoof Googlebot. robots.txt cannot stop them.](https://unsourced.app/research/robots-txt-cant-stop-ai-scrapers) — Nobody blocks Googlebot, so scrapers wear its name and walk in the front door. As publishers gain legal rights over AI data, a self-reported name will not hold up. Proof of identity will. - [Bing now shows AI citations. 4 things it still will not tell you.](https://unsourced.app/research/bing-ai-performance-4-things-it-wont-tell-you) — Bing reports AI citations and four new Copilot views. A real milestone, but the metrics are strictly one engine's slice. Here are four things they leave out — and why each one matters. - [A score is a guess. Ask for the receipt.](https://unsourced.app/research/ai-visibility-score-is-a-guess) — An "AI visibility score" is a modelled guess. Here is what evidence looks like instead, and a simple test you can run on any tool — including ours — to tell measuring from modelling. - [Forensic GEO: an agency retention strategy for AI search](https://unsourced.app/research/forensic-geo-agency-retention) — Rankings hold, AI citations collapse, and the quarterly rank-tracking report cannot see it. The way through is to stop reporting a number and start reporting evidence — forensic GEO as a retention anchor. - [Stop optimising for the machine that is eating you](https://unsourced.app/research/ai-search-stop-optimising-start-auditing) — Your content is the supply chain now, not the traffic source. Optimising harder feeds the machine answering around you more efficiently. The shift is small and total: stop asking how to rank inside the answer, start asking who extracts your work and what you get back. - [Stop blocking AI crawlers. Bait, keep, or block.](https://unsourced.app/research/bait-keep-or-block-ai-crawlers) — Block-everything and allow-everything are both wrong, for the same reason: they treat every bot as identical when they are nothing alike. There are three kinds of AI crawler, not two, and the missing category is the one that matters. - [AI crawlers: how to tell the real ones from impostors](https://unsourced.app/research/ai-crawler-spoofing) — Most analytics tools trust the User-Agent, but that name is a claim, not a credential. Anyone can put GPTBot in a header in one line of code. The fix is not to trust names harder. It is to verify identity at the network layer.