Proofsource review (2026): error bars on every AI visibility number, from $9/mo
Proofsource prices AI tracking per question and per engine from $9/mo and puts a 95% confidence interval on every score. What the meter buys, and where it stops.
Proofsource is an AI visibility tracker from Grovio Labs, Inc., a Delaware company. It is not ProveSource, the social-proof popup tool. Check the spelling before you check the price.
It asks your buyer questions on ChatGPT, Perplexity, Claude and Google AI Overviews, daily or weekly. It reports mention rate, share of voice and position per engine. Every percentage carries a 95% Wilson interval. That last detail is the reason it gets a full review here. Most trackers print "38% visibility" off a handful of samples and let you believe the decimal. Proofsource prints the range and lets you see when a week-on-week change is noise.
Who it suits: small brands and agencies that want to pay only for the questions and engines they track. Who it doesn't suit: anyone who needs more than four engines without a sales call.
What you pay
Prices were read on proofsource.co/pricing and proofsource.co/llms.txt on 20 October 2026.
| Plan | Price | What it includes |
|---|---|---|
| Free trial | Free, 2 days | 25 questions on 4 engines, 200 answers, 100 trial credits, no card |
| Per brand | From $9/mo | $28/mo daily or $7/mo weekly cadence fee, plus a price per question per engine |
| Agency | Free workspace | Unlimited clients, client logins, white-label reports; each brand priced on the card |
| Managed | From $1,000/mo | All 10 engines, fixes and outreach done for you, named strategist, monthly call |
The card is a meter. Each brand pays a cadence fee, then a per-question price on each engine you switch on. Claude costs more per question because, the page says, it searches the web and reads about 14k words per answer. Questions above 50 per brand get 20% off. Yearly billing costs 10 months. Team members are unlimited and there are no overages.
The $9 floor is 5 weekly questions on ChatGPT. That is a price to try the interface, not a price to run a program on.
What works
Confidence intervals. A 95% Wilson interval on mention rate, share of voice and position is the right way to report sampled AI answers. It also tells you when you need more questions before drawing a conclusion.
The AI Traffic view. Connect a Cloudflare zone (one click or an API token) or send access logs as a log drain. Proofsource counts daily requests from GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and others. Each hit is checked against published IP ranges or reverse DNS, and spoofed bots get a label. Pages then sort into three buckets: crawled and cited, crawled but never cited, cited but rarely crawled. That is a crawl-to-citation view, and it is rare at this price.
The fix loop. Proofsource ranks fixes, drafts content for a person to approve, publishes to WordPress and checks back on each fix at days 7, 14 and 28. It also flags wrong facts with the likely source page, lists listicles that leave you out and shows review-site ratings. REST API, MCP access and CSV and JSON exports are there, and you keep your data when you leave.
What to check before paying
The engine count. The pricing table lists 10 engines for every brand. Only four are self-serve. Google AI Mode, Gemini, Copilot, Grok, DeepSeek and Meta AI need a sales call or the Managed tier. Ask which engines your quote actually covers.
GA4. Search Console connects in the app. GA4 is read with Viewer access and set up by the Proofsource team on a call. Once it runs, it reports AI-referred sessions, conversions and revenue by platform. Until then, the conversion half of the story is missing.
Crawler sources. Cloudflare or a log drain. The pages we read name no other CDN integration. On CloudFront or Fastly, ask how the log drain works before you count on it.
Vendor comparisons. Proofsource publishes many comparison pages against other tools. They are its own marketing.
Where it falls short
Proofsource has the right chain of ideas: crawl, citation, click. The gaps are coverage and setup. Four self-serve engines leave out Gemini, Copilot and Google AI Mode, which is a lot of Google to skip. The crawler data depends on Cloudflare or a drain. GA4 needs a call. Prompt research is absent from the pages we read: no search volumes, no difficulty scores, no query fan-outs.
When Promptwatch is the better buy
Promptwatch runs the same chain self-serve and wider. Agent Analytics reads AI crawler logs from Cloudflare, AWS CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN or a custom HTTP source, with a crawl-to-citation path and error tracking. Visitor analytics installs by script or GTM template and attributes AI-referred visits to conversions, with no onboarding call in between. Monitoring covers ChatGPT, Gemini, Claude, Perplexity, Grok, Llama, DeepSeek, Mistral, Copilot and Google AI Overviews and AI Mode. Prompt tracking adds search volumes, difficulty scores and query fan-outs.
The plan to compare is Professional at $245/mo: 150 prompts, 18,000 responses, 25M crawler logs, 1M visitor events, 2 projects and city-level tracking. If you only need tracking and visitor analytics, Essential is $95/mo for 50 prompts across all tracked LLMs. The free Explore plan gives 10 ChatGPT prompts to test with.
Verdict
Proofsource is the most statistically honest tracker at the bottom of the price range, and its crawl-to-citation buckets are a good idea done cheaply. Four engines, Cloudflare-or-drain crawler data and a GA4 setup call keep it a narrow tool. For a team that needs the whole chain on its own schedule, Promptwatch Professional is the better buy.
FAQ
How much does Proofsource cost?
From $9/mo for 5 weekly questions on ChatGPT. Each brand pays $28/mo for daily checks or $7/mo for weekly, plus a price per question per engine. Managed starts at $1,000/mo.
Does Proofsource track AI crawlers?
Yes. It reads crawler requests from a Cloudflare zone or a log drain, verifies each bot and shows which crawled pages get cited.
Which engines are self-serve?
ChatGPT, Perplexity, Claude and Google AI Overviews. The other six need a call or Managed.
See plans at promptwatch.com.