Common questions answered
Everything you need to know about how Unsourced works, what it measures, and what to expect when you start.
Getting started
Your free trial lasts 14 days with full access to all features. No credit card is required to sign up. If you choose to continue after the trial, you can add your payment details in Settings at any point.
Yes — Unsourced has a native WordPress plugin available from the WordPress Plugin Directory (wordpress.org/plugins/unsourced-ai-bot-monitor).
Once connected with your API key from the Settings page, it detects AI bots at the server (PHP) level — catching crawlers that do not execute JavaScript, which a standard JS snippet would miss. It also auto-injects the tracking beacon and Origin Signal into every page, so no code changes are needed.
The plugin is the recommended setup for all WordPress sites, including managed hosting environments like WordPress.com, Kinsta, and WP Engine where nginx log access is unavailable.
The setup takes around five minutes:
- Add your website URL from the dashboard.
- Verify ownership — a meta tag in your HTML or a comment in your robots.txt file.
- Connect bot monitoring — choose the method that fits your stack:
- WordPress — install the Unsourced plugin from the WordPress Plugin Directory (server-side, no JS required).
- Cloudflare Worker — download your pre-configured worker script and deploy it in the Cloudflare dashboard. Works on any site behind Cloudflare — Webflow, Framer, Shopify, custom stacks.
- JS snippet — paste one script tag into your <head>. Works on any platform.
Your first citation scan runs on the next Monday at 04:00 UTC. Pro customers also receive daily scans.
Yes — there are two options for non-WordPress sites:
- Cloudflare Worker (recommended for any site behind Cloudflare) — after verifying your site, download a pre-configured worker script from Settings → Integrations. Deploy it in the Cloudflare Workers dashboard, add a route for your domain, and bot detection and Origin Signal are active with no code changes to your site.
- JS snippet — paste a single script tag into your site's <head>. Works on any platform with no server access required. Catches bots that execute JavaScript; bots that do not execute JS are not captured with this method.
Both options are available in Settings → Integrations after your site is verified.
Unsourced sends prompts to third-party AI providers (OpenAI, Anthropic, Google, xAI, Groq, Perplexity) to run citation checks. These prompts contain your site's URL and topic keywords — not your raw page content.
We do not use your data to train any AI model. Unsourced is a consumer of existing AI models, not a trainer. We query them on your behalf, the same way you would if you typed a question into ChatGPT yourself.
All AI providers we use operate under commercial API agreements. For details on how each handles API data, see the sub-processors section of our Privacy Policy.
Standard plan customers can add up to 5 verified sites. Pro customers can add up to 7. All sites are monitored independently, with separate bot logs, citation reports, and PDF exports for each.
Citation monitoring
Almost certainly not. Zero citations on a first scan — or even after several scans — is the normal state for the majority of independent sites. Most content on the web is not cited by AI models. Unsourced surfaces this reality accurately: if your site is not being mentioned, you will see 0. The scan itself is verified by the system; if it ran, the results are real. A result of zero is useful data — it tells you where you stand so you can act on it.
Unsourced checks seven AI models on every weekly scan:
- ChatGPT (training data) — checks what GPT-4o knows from its training corpus.
- ChatGPT (web search) — uses GPT-4o mini with live search enabled, mirroring what real ChatGPT users see when search is active.
- Claude — checks Anthropic's Claude model.
- Gemini (grounded search) — uses Gemini with live Google Search grounding, reflecting real-time web results.
- Grok (xAI) — checks xAI's Grok model.
- Meta Llama — checks Meta's Llama model via Groq.
- Perplexity — checks Perplexity AI's search-grounded model, which answers queries with live web citations.
Together, these seven models cover both AI training data and the live, AI-augmented search experience your readers are using today.
A citation is recorded when an AI model responds to a prompt with a positive, unreserved mention of your domain or site name. Responses that only contain disclaimers ("I don't have information about this site") or negative qualifications do not count. Each citation is logged with a confidence score, the surrounding response excerpt, and which platform it came from.
The confidence score indicates how strongly the system is able to verify the citation:
- 1.0 — Live crawl confirmed. An AI provider visited your site live during this scan — not from training data.
- 0.85+ — Cited with verbatim phrase matches from your content.
- 0.7 — Cited with contextual match.
- 0.6 — Cited (base confidence — domain or name mentioned positively).
- 0.25–0.45 — Implicit citation: your content phrases appear in the AI's response but your site is not named.
An implicit citation occurs when an AI model's response contains phrases or passages that closely match your published content — but the AI does not name your site as the source. This is detected automatically and flagged separately in your citation feed. It may indicate that your content has been ingested and is influencing AI outputs without attribution.
For Standard plan customers, the citation scan runs every Monday at 04:00 UTC. For Pro customers, an additional daily scan runs at 05:00 UTC (skipping any site already scanned that same Monday). Pro customers can also trigger on-demand scans from the dashboard (4 per calendar month).
Reports
Both reports cover the same underlying data. The difference is framing and purpose:
- AI Visibility Audit — structured for content and SEO strategy. Sections cover your citation rate by platform, competitor domains appearing in AI responses, topic performance, and the gap between bots that crawled you and the citations that followed. Includes AI-generated strategic recommendations.
- Evidence Report — structured for legal or licensing contexts. Sections cover bot identity and access logs, page-level crawl evidence, citation results with response excerpts, named competitor domains, verbatim phrase matches, and recommended next steps.
Evidence Reports are provided as structured data to support review, complaint, licensing discussion, or legal escalation where appropriate. They are not a legal determination and do not constitute legal advice.
Yes. From the Reports page, click "Share" to generate a public link valid for 30 days. Recipients can view a summary page and download the PDF without logging in. The shared link can be sent to a solicitor, a licensing contact, or anyone else who needs to review the data.
When generating a report, you can select 30, 90, or 180 days of data. Standard customers have access to 6 months of history; the data is always available within that window. Trial customers have the same access during the 14-day trial period.
Bot monitoring
It's not all-or-nothing — AI crawlers fall into two camps, and it helps to know which is which:
- Harvesters — AI networks that scrape your content with no intent to cite or attribute you. If you want to block these, you can do it yourself in Cloudflare's bot controls or via robots.txt.
- Citers — the models that actually mention your site and drive visibility — the ones worth tracking. Origin Signal on your public pages verifies which AI models cite you live.
The catch: you can't tell which bot is which until you measure. That's what Unsourced is for — it shows you who's visiting, who's citing, and who's just taking, so you can decide who to allow and who to block in your own tools, with evidence instead of guesswork. Unsourced monitors and informs; the decision stays yours.
Origin Signal tells you whether a citation came from an AI that visited your site live during the scan, or from its training data. Live-crawl citations are timestamped and included in your Evidence Report. Training data citations are also recorded — so you have the full picture either way.
The WordPress plugin installs directly from the WordPress Plugin Directory (wordpress.org/plugins/unsourced-ai-bot-monitor). Once connected with your API key from the Settings page, it intercepts incoming requests at the server level and sends bot detection events to Unsourced automatically. It also injects your site's tracking beacon into every page, so beacon-based detection and Origin Signals work without any code changes. This approach captures AI bots that a standard JavaScript beacon cannot — bots that do not execute JavaScript.
The Cloudflare Worker runs at the network edge — before requests reach your server. When an AI crawler visits your site, the worker detects it by user-agent signature, reports the event to Unsourced in the background (using ctx.waitUntil so there is zero latency impact), and lets the original request through unchanged.
It also enables Origin Signal automatically — no code changes needed on your site.
To set it up: go to Settings → Integrations → Cloudflare Worker, download the pre-configured script for your site (credentials are baked in — no environment variables needed), then deploy it in the Cloudflare Workers dashboard and add a route for your domain. Works on Webflow, Framer, Shopify, Ghost, custom stacks — anything behind Cloudflare.
Unsourced detects 20 identified AI bots by user-agent signature, including GPTBot, ChatGPT-User (OpenAI), ClaudeBot, anthropic-ai (Anthropic), Google-Extended, GoogleOther, PerplexityBot, CCBot (Common Crawl), Bytespider (ByteDance/TikTok), Meta-ExternalAgent, and others. New bot signatures are added as they are documented publicly by their operators.
Yes. The WordPress plugin is the recommended solution for managed WordPress environments (WordPress.com, Kinsta, WP Engine, etc.) where you do not have access to raw server logs. The plugin operates at the PHP level inside WordPress itself, so no server access is required.
Plans and billing
- Standard (£29/month) — weekly citation scans, up to 50 seed pages per scan, up to 5 sites, weekly digest email.
- Pro (£49/month) — everything in Standard, plus: daily citation scans, 4 on-demand scans per month, up to 200 seed pages per scan, up to 7 sites, automated citation alerts, and webhook delivery.
A higher cap means broader topic coverage and more varied prompts per scan.
Yes. You can cancel your subscription at any time from the Settings page via the Stripe customer portal. Cancellation takes effect at the end of your current billing period. You retain full access until then. There are no cancellation fees or lock-in periods.
If you experience a technical issue that prevents you from using the service, please contact rene@unsourced.app and we will work to resolve it or provide a fair remedy. We do not offer refunds simply because citation results were lower than expected — zero citations is a valid, accurate result for many sites, not a product failure.
Data and privacy
No. Unsourced doesn't train AI models — we're a consumer of existing models, not a trainer. Your bot logs and citation results are stored only in your account and are never sent to any AI provider. To check citations, we send short generated questions to AI providers through their business APIs; under those providers' API terms, data sent this way is not used to train their models. Full details and our sub-processor list are in our Privacy Policy.
Your account and monitoring data are stored on servers in the European Economic Area (Hetzner, Germany). To check AI citations we also send generated questions to third-party AI providers — some based outside the UK/EEA — under Standard Contractual Clauses (SCCs). Unsourced is registered with the UK Information Commissioner's Office (ICO, registration ZC150902) and processes data in accordance with UK GDPR. Our full sub-processor list and transfer safeguards are in our Privacy Policy.
Yes. You can permanently delete your account and all associated data from the Settings page. This is irreversible. All data — sites, crawl events, citation events, reports, and account information — is permanently removed in accordance with your rights under UK GDPR Article 17.
Still have a question?
Get in touch directly — there's a real person on the other end.