Applitools vs Percy: visual regression pricing and coverage.

What visual regression testing actually solves

Visual regression diffing catches a class of bugs that functional assertions usually miss: a font that renders slightly wrong, a layout that breaks on a particular viewport, a colour that drifts after a CSS refactor, an image that fails to load on one path. Pure functional tests pass because the underlying DOM is correct; the customer experience nevertheless degrades. Visual regression makes the degradation testable.

Both Applitools and Percy slot into an existing Playwright, Cypress, or Selenium suite via a small SDK. A test makes a snapshot call (eyes.check() for Applitools, percySnapshot() for Percy), the screenshot is uploaded to the vendor platform, the platform diffs it against the baseline, and a PR comment or dashboard surface flags the diff for human review.

Diff technology

Applitools Eyes uses what the vendor calls Visual AI, an analysis layer above raw pixel diffing that attempts to distinguish meaningful changes (a text label moved by 50 pixels) from cosmetic noise (a one-pixel anti-aliasing difference). The published documentation describes layered analysis with classifications for content, layout, style, and movement. The Visual AI claim is the published differentiator against simpler pixel-diff approaches and is the basis for Applitools' per-checkpoint pricing model (a more complex analysis warrants the unit cost).

Percy takes a similar approach with pixel-level diffing plus AI-classified changes, hosted on the BrowserStack platform (browserstack.com/percy). The published features include responsive snapshot capture across viewports and DOM-based diffing for animation-heavy interfaces. The AI-classification layer is comparable in scope to Applitools' without being identical in published taxonomy.

Neither vendor publishes an apples-to-apples false-positive benchmark against the other. Customer-reported anecdotes vary by application; teams running animation-heavy SPAs report different false-positive profiles than teams running mostly static content sites. The honest test is the team's own pilot on a representative set of pages.

Pricing unit

Applitools prices by checkpoint, where a checkpoint is a single visual verification against a single configuration. A test that takes 10 snapshots across 5 browsers and 3 viewports consumes 150 checkpoints. The published tiers scale with checkpoint volume; enterprise contracts are custom.

Percy prices by snapshot, with the published tiers on the Percy pricing page scaling with snapshot volume per month. The unit is simpler: one snapshot is one snapshot, with multi-browser comparisons counting separately. For a team that runs the same test across many configurations, Percy's simpler unit often produces a lower bill at equivalent coverage, but only at equivalent coverage; teams should normalise on coverage, not on raw unit count.

Framework support

Both vendors support the major end-to-end frameworks (Playwright, Cypress, Selenium, WebdriverIO) and have integrations for storybook-driven design system workflows. Both publish CI integrations for GitHub Actions, GitLab CI, CircleCI, and Jenkins. Neither has a meaningful framework gap relative to the other in 2026.

Storybook context

Both products are usable with Storybook but a third product, Chromatic, is purpose-built around Storybook and ships from the same team that maintains Storybook itself. For a design-system workflow centred on Storybook, the cleaner comparison is Chromatic against either Applitools or Percy. See the visual regression entry on the category overview for the broader vendor map.

Where each fits

Applitools fits when the team values the published Visual AI classification, when the procurement context already includes broader Applitools platform products (Ultrafast Grid, Eyes Component), and when checkpoint-based pricing aligns with the test profile (low snapshot count, multi-dimensional coverage).

Percy fits when the team is already in the BrowserStack ecosystem, when snapshot-volume pricing aligns better with the test profile (high snapshot count, narrower configuration coverage), and when the BrowserStack bundle is attractive for the broader testing stack.

Operational considerations

Both products require discipline around baselines. The first time a test runs, the platform records a baseline; future runs compare against that baseline. Teams that ignore the baseline-management workflow end up with stale baselines that drift away from the intended design, defeating the purpose of the diff. Both vendors publish baseline-management guidance; following it is operational labour that does not go away with the platform choice.

False positives are the dominant operational cost. Every diff a human has to review and accept or reject is engineering time. Both vendors' AI classification reduces this load but does not eliminate it. Budgeting reviewer time as part of total cost (often 0.1 to 0.3 FTE for a team running visual regression on every PR) is a more honest cost frame than list price alone.

How to choose

Run a one-week parallel pilot on the same test suite under both vendors. Count true positives, false positives, baseline-creation effort, and PR-review overhead. Calculate the projected monthly bill at current snapshot volume. The vendor that wins on (correct diffs × reviewer hours saved) net of license cost wins the procurement decision. The pilot is the only way to make this call honestly.

Frequently asked questions

Why is Applitools more expensive at scale?

Applitools' pricing is tied to checkpoint volume (each screenshot that the platform diffs counts as one or more checkpoints depending on viewport and configuration). A test suite that takes many screenshots across many browsers consumes checkpoints quickly. Percy's pricing is also volume-based but the unit is the visual snapshot rather than the multi-dimensional checkpoint, which can produce a lower bill for the same coverage at scale.

Does Percy still exist after the BrowserStack acquisition?

Yes. Percy is sold and supported by BrowserStack and remains an active product. The vendor page now sits under the browserstack.com domain and pricing is published on the Percy pricing page. Buyers should note that the broader BrowserStack platform context affects bundled deals; standalone Percy procurement is still supported.

Are AI-described differences trustworthy?

Both vendors publish AI-described diffs that classify changes as content, layout, style, or moved-element. The classifications are useful for triage but should not be treated as ground truth; a human reviewer remains responsible for accepting or rejecting each diff, particularly on customer-facing flows.

Which one fits Storybook better?

Both have first-party Storybook integration. Chromatic is the third option specifically built around Storybook and is owned by the same team behind Storybook itself; that integration depth is hard to beat. For a Storybook-led design system workflow, the right comparison is Chromatic against either Applitools or Percy, not Applitools against Percy directly.

Can I use either without Playwright or Cypress?

Yes. Both have direct browser SDKs and CLI invocations for standalone use. The common usage pattern, though, is to call the visual snapshot API from inside an existing functional test framework so that the same test serves both a behavioural assertion and a visual diff.