ZeroRank review (2026): how it scores on the four jobs of an AI search program
We review ZeroRank against prompt research, citations, AI crawler access and AI-referred traffic. Strong research and citation views, a readiness check instead of crawler logs, and AI traffic read through Google Analytics.
ZeroRank used to be a simple tracker. After a busy summer of releases and a repricing, it now looks more like a full AI search workbench, from $39 a month. That made it worth a proper review.
We only work on AI search. No link building, no local packs. Every program we run comes down to four jobs: choose the right prompts, understand the citations behind each answer, make sure AI crawlers can reach the pages that should be cited, and count the visits that come back. We scored ZeroRank on those four.
Sources: zerorank.ai/pricing, zerorank.ai, the About page, the platform pages for AI search optimization, agentic workflows and prompt research, and the changelog (latest entry 19 September 2026), all read on 28 October 2026. There is no llms.txt; that URL returns a 404. Promptwatch facts are from its llms.txt, re-checked on 21 October 2026.
Scorecard
| Job | ZeroRank | Our read |
|---|---|---|
| Prompt research | Strong | AI volume estimates, Google volume, trend, Search Console signals |
| Citations | Strong | Domain types, UGC tab, full answer text stored |
| AI crawler access | Partial | Readiness check, no published server-side logs |
| AI-referred traffic | Partial | Google Analytics based, which ZeroRank calls a lower bound |
Prompt research: better than we expected
Prompt Research shipped in July. Give it a topic, a competitor's domain or a question and it returns prompt ideas with an estimated monthly AI search volume, Google search volume where that is measured, a 12-month trend and a country picker. Connect Search Console and it adds those signals too. An AI Volume column sits on the prompts table itself, so prioritising does not mean exporting.
Tracking is flexible. Each prompt has its own cadence: daily, weekly, monthly or manual. The pricing page lists 17 AI platforms on every plan, naming ChatGPT, Perplexity, Gemini, Google AI Overviews, Claude, Copilot, Grok and DeepSeek.
The limit is the answer budget. Starter's 2,400 answers a month cover, by ZeroRank's own example, 26 prompts a day on three standard models. Growth ($199) covers 150, Agency ($499) 350. Higher-cost models spend more. Plan the prompt set and the engine mix together, or the budget makes the choice for you.
Citations: a good working view
ZeroRank reads each answer back with its sources and keeps the text. The Chats view shows what every model said, word for word. Sources breaks citations down by domain and by URL and labels each domain as Corporate, Editorial, UGC, Reference, Institutional or Other. A separate UGC tab and a gap analysis help you find the forums and review pages that carry a competitor and not your client.
Promptwatch, the platform we run client programs on, covers the same ground and goes further on time. Its citation analytics include Reddit and YouTube citations, offsite mentions and citation type breakdowns, and citation trends show how sources and volume shift over months. When a client asks what changed since last quarter, the trend view is the chart we open.
AI crawler access: a readiness check, not logs
This job matters most in AI search and gets the least attention. A page cannot be cited if the crawler never gets a clean copy. Soft 404s, blocked paths after a redesign, JavaScript-only content, bot rules in the CDN that nobody remembers adding. Any one of them can quietly undo months of content work.
ZeroRank's technical analysis, labelled Agent Readiness and in beta, checks robots.txt, llms.txt, accessible markup, PageSpeed and what each AI crawler actually got back from a page. That is useful for a quick audit. It is still a test you run, not a record of what happened. It cannot tell you, for example, that PerplexityBot hit 40 URLs on Tuesday and a third of them errored.
ZeroRank does not publish server-side crawler logs. Promptwatch's Agent Analytics does: real-time AI crawler logs from Cloudflare, AWS CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN or a custom HTTP endpoint, with error tracking and a path from crawl to citation. It starts on Professional ($245/mo) and is on every agency plan. For our work, that is the difference between guessing why a page was dropped and knowing.
AI-referred traffic: honest about its limits
AI Traffic arrived in July. Connect Google Analytics and ZeroRank shows sessions, visitors, engagement and conversions from ChatGPT, Perplexity, Gemini, Claude, Copilot, Grok, DeepSeek, Mistral, Meta AI and You.com, as a funnel from seen in AI answers to visited to converted. Landing pages carry citation counts. There is an API endpoint for it.
ZeroRank is upfront that Google files AI Overviews clicks as organic and Copilot referrals as bing.com, so the figure is a lower bound, and GA data can lag up to 48 hours. We respect that.
First-party tracking matters for a different reason. A client's GA property is often half-configured, with filters, consent quirks and channel groupings someone set up in 2021. Promptwatch's visitor analytics run on its own lightweight script or GTM template, count AI-referred visits and conversions, and sit next to the prompts and crawler logs in one workspace. It does not change how Google labels an AI Overviews click. It does mean the AI traffic number does not inherit someone else's GA settings.
The fix layer, briefly
ZeroRank also acts. Recommendations are ranked by how many tracked prompts they touch. Copilot drafts a change as a diff against the live page. Agentic Workflows fire on triggers, such as an answer dropping you, and can pull the competing sources, brief the team in Slack and push an approved change to WordPress or Webflow. Nothing goes live without approval, and workflows are on every paid plan. That is good value at this price.
Who ZeroRank suits
A small in-house team that wants broad engine coverage, good prompt research and an approval-gated content loop on a small budget, and can accept a readiness check and GA-based traffic as its diagnosis layer.
How we run it instead
Our programs run on Promptwatch, because crawler logs and first-party attribution are the two jobs where ZeroRank stops at a proxy. When we can see the bot visit, the citation and the session in one place, the monthly report explains itself.