Promptwatch vs Trakkr vs LLM Pulse vs Rankscale: Features Compared
How we compare crawler analytics, citation types, MCP, and publish features across Trakkr, LLM Pulse, Rankscale, and the Promptwatch stack we actually run.
Clients paste Trakkr, LLM Pulse, and Rankscale into Slack because the homepages now say crawlers, MCP, and traffic. The marketing caught up to the category, and that is good for buyers and confusing for procurement. We still run Promptwatch. This is the feature checklist we use on the call, not a price sort. Price is the last column, because the wrong feature set at any price is still the wrong feature set.
The reason the Slack paste happens is that the words on these homepages started to converge. A year ago, a prompt tracker said "prompts" and a crawler tool said "crawlers" and the two did not overlap. Now every prompt tracker's homepage mentions crawlers, MCP, and traffic, because those are the words buyers search for. The products underneath the words did not all converge to match. Some did the work and some put the words on the page. The checklist below is how we tell which is which, and it is why price is the last column. A cheap tool that does not do the work is not a deal. It is a line item we will cancel in month two.
Google's AI features pages stay the Overviews rulebook. Anything below that is a tool choice, not a Google rule.
The Google doc is the one piece of the stack that is not a tool choice. AI Overviews and AI Mode are Google surfaces, and Google tells you how it ranks them. Everything else in this comparison, ChatGPT, Perplexity, Claude, Gemini, is a surface Google does not control and does not document. The tool choice is how you measure those surfaces, and the tool choice is ours, not Google's.
What we need in a retainer, module by module
Prompts. We write them with the client. We need fan-outs, personas, and city splits when the brand is not "global average." Promptwatch has state and city targeting on Professional and the agency plans, while Essential only has country-level targeting. Rankscale has credits, which is a different model that puts the burden on you to watch the meter. LLM Pulse Starter is 50 prompts on ChatGPT, Perplexity, Gemini, AI Mode, and Overviews, with Claude gated to Enterprise. Trakkr Growth is 50 prompts across eight models, daily, then prompt packs at $39 to $99. Fifty prompts is a workshop, not a roster, and a team that fills it in week one will be buying packs by month two.
The prompt module is where the four tools diverge first, because the prompt count is the unit that runs out. Fifty prompts sounds like a lot until you build a fan-out. A single buyer query, "best CRM for mid-market," fans out into five or six phrasings the moment you take it seriously. "Top CRM for mid-size companies," "leading CRM for mid-market," "what CRM do mid-market teams use," and so on. One buyer question is six prompts. Fifty prompts is eight buyer questions, which is a workshop, not a program. The team that fills the fifty in the first week is the team that buys packs in the second month, and the pack price is the real price of the entry plan.
Citations. The QBR question we ask is simple: was it our URL, a competitor URL, Reddit, or YouTube? Promptwatch stores those types and the trend behind them, which is what lets us tell a client whether a win is holding or slipping. Trakkr's site is good on "source behind the answer" and on who displaced you, which is a useful framing for a competitive read. LLM Pulse lists citations, share of voice, and owned-media tracking. Rankscale is scoring pages, not filing citation types, so it answers "how good is this page" rather than "where did the mention come from."
The citation module is where the four tools answer different questions, and the question matters more than the feature name. Promptwatch answers "where did the mention come from," which is the question that lets you repeat a win. Trakkr answers "who displaced you," which is the question that lets you read a loss. LLM Pulse answers "what is your share of voice," which is the question that lets you read a market. Rankscale answers "how good is this page," which is a page-quality question and not a citation question at all. A team that buys Rankscale for citations buys a page scorer and then wonders why the mention count is missing. The feature names sound similar. The questions are not.
Crawlers. This is where Trakkr got interesting. Growth includes crawler analytics, and their docs talk CDN and CMS ingest and a crawl-cites-click path. On Promptwatch Professional or an agency plan, Agent Analytics publishes ChatGPTBot, ClaudeBot, PerplexityBot, GoogleOther, and Meta, plus the CDN list we connect, which is Cloudflare, CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, and custom HTTP, with errors on the path. Essential has no listed crawler-log allowance. LLM Pulse and Rankscale do not publish this. We will not diagnose a miss as "thin content" if we never saw a fetch, because the cause might be a 403 the bot hit before any content was read.
The crawler module is the one that changed how we diagnose misses. Before crawler logs, a miss was "thin content," because that was the only diagnosis available without a fetch record. After crawler logs, a miss splits into two cases. Case one is the crawler fetched the page and the content did not earn the citation, which is a content problem. Case two is the crawler never fetched the page, or hit a 403 on the way in, which is a technical problem. The two cases have different fixes, and the fix for case two is not to write more content. The fix for case two is to unblock the bot. Without the fetch record, a team writes more content for a page the crawler cannot reach, and the miss does not move. The fetch record is what stops that waste.
Traffic. Promptwatch visitor analytics, through a script or GTM, is what goes on the monthly report next to mentions. LLM Pulse has AI traffic analytics. Rankscale will not attribute revenue, and their own docs say the lag runs 4 to 8 weeks through a proxy. Trakkr shows an AI traffic tile in marketing, but we have not treated that as conversion reporting, because a tile and a conversion path are different claims.
The traffic module is the one that separates a mention report from a pipeline report. A mention report says the brand was visible. A pipeline report says the visibility drove a session, and the session did something. The conversion path is the part that makes a mention worth budget, because a mention that drives no session is a brand impression, and a brand impression is hard to defend in a pipeline review. The Rankscale lag of 4 to 8 weeks is the part that disqualifies it for monthly reporting, because a number that lands two months late is a number that arrives after the decision it was meant to inform. The Trakkr tile is the part we do not treat as a conversion claim, because a marketing tile is not a measured path.
Publish. Promptwatch Content Agents write into Webflow or Framer with a human review inbox, so the page does not ship until a person accepts it. Trakkr describes automations and apply-to-site. Until a client shows us that writing into their CMS with review, we do not call it the same feature, because the review step is the part that keeps the page from being a liability. LLM Pulse has a GEO Writer. Rankscale produces fix lists whose audit scores the vendor says move to a 30-day average, which is a slow signal for a fast-moving channel.
The publish module is the one that turns measurement into output, and the review step is the part that makes it safe. A content agent that writes straight to the CMS is a liability, because AI content that ships without a human pass is content that can be wrong, off-brand, or duplicative, and the wrongness ships to a live page. The review inbox is the gate that catches the wrongness before it ships. Trakkr's "apply-to-site" language is the part we want to see demoed, because the language could mean a reviewed apply or an unreviewed apply, and the two are different features with the same name. We do not call them equal until we see the review step.
Access. We need MCP or an API when a strategist lives in Claude. Pulse has MCP on every plan and REST at €299. Trakkr has MCP on Growth. Promptwatch has MCP and REST API from Essential, with Data Studio listed on Professional and the agency plans. Slack and Google Search Console are separate connections. Rankscale API runs $385.
The access module is the one that decides whether the strategist can live in the tool or has to context-switch into it. A strategist who lives in Claude wants MCP, because MCP lets the tool show up inside the conversation instead of forcing a tab switch. The REST API is the heavier integration path, for the engineer who builds a pipeline. The two are not substitutes. MCP is for the strategist, REST is for the engineer, and a tool that has both serves both seats. Rankscale's $385 API is the one that prices the engineer out of the entry plan, which is fine for a team that has the budget and a blocker for a team that does not.
What we will not do
We will not start a retainer on Rankscale $20. Credit burn by engine is how that invoice becomes a surprise, and the surprise lands in the month a stakeholder already asked for a clean number. We will not let a Trakkr trial auto-charge at $100 without a calendar invite on day 13, because the trial that converts without a decision is not a trial. We will not report LLM Pulse weekly numbers as if they were daily, because the cadence is the claim and mislabeling it changes what the client thinks they are buying.
The three things we will not do are the three ways a retainer goes sideways in the first quarter. The Rankscale credit surprise is a budget problem. The Trakkr trial auto-charge is a process problem. The LLM Pulse cadence mislabel is a reporting problem. Each one is preventable with a discipline that costs nothing, and each one is expensive to fix after it happens. The calendar invite on day 13 is the cheapest insurance in the list. The cadence label on the report is the cheapest honesty in the list. The credit watch is the cheapest budget hygiene in the list. None of these require a different tool. They require a different habit.
Kick-off on Promptwatch is $199 a month with crawler logs, state and city targeting, Data Studio, custom reports, and unlimited projects and prompts. That is the agency math against Trakkr add-ons, which run extra brand at $39, AI pages at $79, and white-label at $49. Promptwatch white-label is Enterprise or Custom, not a self-serve Kick-off feature, so do not assume the $199 agency plan ships it.
The agency math is the part that decides which tool a multi-brand team lands on. Kick-off at $199 with unlimited projects and prompts is a flat cost. The Trakkr add-ons stack, because each extra brand, each AI page pack, and the white-label tier are separate line items, and a multi-brand team hits all three. The stack is how a $100 base becomes a $300 invoice, and the stack is why the flat $199 wins the multi-brand math. The white-label caveat is the one that catches a buyer who reads "agency plan" and assumes white-label is included. It is not. White-label is Enterprise, and a team that needs it has a different conversation than a team that needs Kick-off.
FAQ
Would we run Trakkr crawler analytics next to Promptwatch?
Only if a client already bought it. We will not dual-log bots for fun. One fetch story.
Is LLM Pulse enough for a small EU brand?
As a weekly five-engine board with unlimited seats, maybe. As the system we put a marketing lead into, no, because Claude and named-bot logs are missing.
Where do we send them?
promptwatch.com. Keep Search Console for Overviews.