CiteGraph Review 2026

Key takeaways

• CiteGraph's core hook is genuinely different: instead of a visibility score, it shows you the verbatim AI answer you lost, the competitor who won it, and a drafted sentence that could win it back.
• It's a monitoring-plus-drafting tool, not a full optimization platform. Compared to Promptwatch, it lacks AI crawler logs, visitor analytics (actual AI traffic and conversions), Reddit and YouTube citation tracking, ChatGPT Shopping and Ads tracking, prompt volumes and difficulty scores, CMS publishing, and it covers only 4 engines versus Promptwatch's 10+.
• Pricing is unusually honest and cheap for the category: $12 for two one-off scans, then plans from $29/mo. No free tier, because every scan burns real API money.
• The statistical rigor is a real strength. Every rate ships with a margin of error, and the tool refuses to call something a "move" unless it exceeds that margin. Most competitors don't do this.
• It's a very new, bootstrapped product with sparse third-party reviews, English-only questions, and no region or language modeling. Fine for a founder or small marketing team; thin for an enterprise program.

What CiteGraph actually does

Most AI visibility tools give you a dashboard with a score that goes up or down and leaves you to figure out why. CiteGraph starts one step later in the funnel: it finds the specific buyer questions where ChatGPT, Claude, Gemini, or Perplexity named a competitor instead of you, then shows you the exact answer text, the pages the engine cited to build that answer, and a drafted "missing sentence" that would have qualified your product for inclusion.

That last part is the interesting bit. Each diagnosed loss becomes a "play" (a paragraph, comparison row, or pitch), drafted by one AI model and judged by a second one, with unverifiable claims marked "[FILL IN]" rather than hallucinated. It's a smart guardrail, and in my time with the tool the drafts were more restrained than what you'd get from a generic content generator.

The methodology page deserves credit too. Questions are run 3 to 5 times per engine per scan (120 to 400 live API calls), through official grounded-search APIs rather than scraped chat UIs. Every rate is published with a Wilson 95% confidence interval, and a change only counts as movement when it exceeds the combined margins of both scans. In a market where plenty of tools report single-run numbers as fact, this is a genuinely refreshing stance.

The "named vs. cited" distinction is another differentiator. Being named means the assistant mentioned your product in prose; being cited means your own page was used as a source. CiteGraph's own research found that 49% of pages cited when AI recommends a product belong to someone else, which means you can be recommended constantly while your own words never reach the buyer. The cited-source map grades every page that engines pull from as SOLID, THIN, or NOISE based on how substantively it describes you, and flags rival-owned gatekeeper pages. That's actionable in a way a share-of-voice percentage isn't.

Pricing

Current live pricing (the site previously listed higher tiers, which have been retired): Starter at $29/mo for 2 domains, 10 questions per project, 3 runs per question, 6 rivals, 1 seat. Growth at $69/mo (the most picked tier) adds 5 runs, 10 rivals, 5 seats, fact sheets for writers, and CSV export. Scale at $149/mo covers 6 domains with unlimited seats, priority scan queue, and a white-label public dossier link aimed at agencies.

The entry path is the clever part: your first scan is free (one per account), and if you don't want a subscription, $12 buys two full scans, then $12 per scan after that. The 7-day trial on paid plans requires a card but charges nothing up front. There's no ongoing free tier, which the company justifies bluntly: every scan spends real money on live engine calls. Fair, though it does mean the "try before you buy" window is a single scan.

For context, that undercuts Otterly.AI's entry tier ($29/mo) while matching it, and comes in well below Profound's Starter plan at $99/mo billed yearly. If you just want a one-time diagnostic of where you're invisible, $12 is hard to argue with.

Who it's for

The sweet spot is a SaaS founder or a small marketing team that has heard "ChatGPT recommends our competitor" from a customer and wants receipts. The verbatim lost answers and drafted plays map directly onto that conversation. The Scale plan gestures at agencies managing a handful of client domains, and the white-label dossier link helps, but with only 6 projects max and no client portal infrastructure, larger agencies will hit the ceiling fast.

Strengths and limitations

Strengths:

• Verbatim lost answers with the exact sentence that took your place. Nobody else leads with this, and it's the fastest path from "we're invisible" to "here's the specific fix."
• Statistical honesty. Sampled runs, printed error bars, and a refusal to flag noise as movement. The company also publicly states that llms.txt has no measurable effect on citations, even though competitors sell generators for it.
• The public category leaderboards are free, require no signup, and can't be bought. CiteGraph doesn't appear in its own AEO-tools leaderboard and publishes that fact anyway. That's either admirable or a warning sign depending on your read; I lean admirable.
• Cheap, low-commitment entry via pay-per-scan.

Limitations, and this is where the comparison with Promptwatch matters. CiteGraph tracks 4 engines (ChatGPT, Claude, Gemini, Perplexity). Promptwatch monitors 10+ including Grok, DeepSeek, Copilot, Mistral, Meta Llama, Google AI Overviews, and Google AI Mode, which is where a growing share of AI answers actually happen. CiteGraph has no AI crawler logs, so you can't see when bots visit your site, what they read, or whether they hit errors. There's no visitor analytics, meaning no attribution of actual traffic and conversions from AI platforms; you're working entirely from mentions and citations. No Reddit or YouTube citation tracking, no ChatGPT Shopping or Ads Radar, no prompt volume or difficulty data, no query fan-outs, and no multi-language or multi-region monitoring (questions are English-only for now). The drafted plays are helpful, but there's no CMS publishing or automated content agent to close the loop; you take the draft and do the work elsewhere. Integrations are essentially absent: no Slack, no Looker Studio, no Google Search Console, no API on standard plans.

Also worth naming plainly: this is a very early product. Third-party reviews are limited to a handful of brief comments on launch platforms (4.8/5 from 4 votes on MicroLaunch). No G2 or Capterra presence. The founder is actively experimenting with pricing and paywalls in public. None of that makes it bad, but it does mean you're betting on a small team's momentum.

The bottom line

CiteGraph does one thing unusually well: it turns AI invisibility into a specific, readable, fixable problem. The verbatim lost answers, the named-versus-cited split, and the statistically honest movement detection put it ahead of basic prompt trackers on diagnostic depth, and the $12 scan is one of the cheapest credible entry points in the category.

But it's a diagnostic tool with drafting assistance, not an optimization platform. If you need to know why you're invisible, where AI crawls your site, what AI traffic actually converts, how you're doing across Grok, AI Overviews, and AI Mode, and you want agents that plan, write, and publish the fixes, Promptwatch covers that full stack with crawler logs, visitor analytics, Reddit and YouTube tracking, prompt volumes, and CMS publishing. My honest recommendation: use CiteGraph's free scan or $12 dossier if you want a sharp one-time read on your blind spots. Choose Promptwatch if AI search is a channel you intend to grow revenue from, not just measure.

Promptwatch

AI search visibility and optimization platform

View more