Methodology: 40+ public complaints, HN threads, TechCrunch coverage, and Reddit discussion spanning June 2025 through April 2026. Analyzed with RetentionCheck.

The Churn Health Score

Cursor scored 42/100, grade D. The score reflects 1 critical, 2 high, 2 medium, and 1 low severity driver across the public corpus.

For a tool with a $9B valuation, a D grade is jarring. The score does not say Cursor is failing. It says Cursor's loyal customers feel something specific is broken, and they have started to act on it.

The 5 churn patterns

1. June 2025 pricing restructure (critical, 90% confidence)

On June 16, 2025, Cursor moved the Pro tier from 500 fast responses plus unlimited slow responses to $20 of usage billed at API rates. Effective Pro capacity for heavy users dropped roughly 50%, from about 500 requests to 225. Backlash was severe enough that the CEO publicly apologized and refunds were issued.

"I went back to Copilot after the June credit change. Cursor Pro went from 500 to about 225 effective requests at $20 and Windsurf gives me comparable context at $15." Reddit, aggregated

Pricing changes that reduce perceived value at the same price point read as bait-and-switch. The damage is rarely about the dollars. It is about the implicit contract being rewritten.

2. AI support hallucinated lockout policy (high, 85% confidence)

In April 2025, Cursor's AI customer support fabricated a non-existent lockout policy and told users they were blocked from multi-device login. Users canceled based on the false information before Cursor clarified the policy.

"Cursor IDE support hallucinates lockout policy, causes user cancellations." Hacker News thread title

AI-only support without a human gate on policy or billing claims is a trust risk. The dollar savings on human reps are almost never worth the cost of a churn-driving false answer.

3. Opaque credit counter anxiety (high, 80% confidence)

Post-restructure, users describe the credit counter as anxiety-inducing. Usage tracking lags actual spend. Developers cannot predict when they will hit the limit. For a tool engineers use 8 hours a day, the uncertainty is more damaging than the price.

"The credit counter is anxiety-inducing and opaque." Reddit

Flat-rate predictability is a feature. Usage-based pricing without real-time clarity is a churn engine.

4. Cursor 3 agent-first redesign polarized the community (medium, 68% confidence)

April 2026 shipped Cursor 3 with an agent-first interface, moving beyond the traditional IDE model. Developers who originally chose Cursor for its IDE depth are evaluating whether the new direction still serves their workflow. Per InfoQ, community reaction split along that line.

Forcing a redesign on loyal users is a high-risk move when the prior version is the reason they chose you.

5. Claude Code and Windsurf gaining migration velocity (medium, 72% confidence)

Alternatives have absorbed real migration. One Reddit commenter switched from Cursor at "$2k a week on premium models" to Claude Code Max and reported being "equally as prolific and paying 1/10th the price." Windsurf is cited as the "easiest transition" at $15 per month.

The lesson: pricing parity games are losing games. Cursor's moat was always depth of IDE integration. Competing on lower per-token price erases that moat.

What Cursor does well

Reviews and Hacker News comments consistently praise Cursor's IDE depth, codebase-wide context, and integration polish. Developers who stay cite the raw productivity boost of multi-file context-aware AI. At its best, Cursor is category-defining, and the $9B+ valuation reflects real product-market fit on the core use case.

The churn story is not Cursor is bad. It is Cursor's monetization and trust hygiene are eroding the loyalty of the customers who made it famous.

3 things Cursor could fix

Publish a 12-month Pro pricing commitment. Grandfather existing customers on credit allowances. Separate a power-user tier with pay-as-you-go so base Pro stays predictable.
Add a human gate to AI support on policy and billing claims. AI answers questions, humans confirm anything about access, money, or accounts.
Real-time credit display with cost estimates per action. Soft-limit warnings before cutoff. User-controlled monthly caps.

What this means for your SaaS

If you take one thing from the Cursor story, take this: the first pricing change you ship is the one your earliest customers will judge you by forever. The dollars matter less than the perceived rewrite of the deal.

The same pattern shows up in mid-stage SaaS regularly. A founder adds a metered tier, removes the word unlimited, or moves a feature to a higher plan, and the cohort that loved them most produces the loudest exit signal. The math is rarely the issue. The feeling of a broken contract is.

If you have cancellation feedback sitting in a spreadsheet or support inbox, you can run your own analysis in 30 seconds. Paste your data, get severity-ranked churn insights with verbatim quotes. Free tier covers 3 analyses per month.

You can also browse our example analyses to see what the output looks like before you paste anything.

Key takeaways

Perceived bait-and-switch is the most expensive form of churn. It taxes every future feature launch too.
AI-only support without a human gate on policy questions is a trust gamble. The viral bad answer costs more than a year of human rep payroll.
Usage-based pricing without real-time clarity is a churn engine. Pair metered pricing with predictability tools.

Brian Farello is the founder of RetentionCheck, an AI-powered churn analysis tool for SaaS teams.

Why Cursor Users Are Canceling: I Analyzed 40+ Public Reviews