Visual regression testing tools compared

The main options are the hosted services Chromatic, Percy, Argos, Applitools, and UI Verify, plus the free toHaveScreenshot comparison built into Playwright. All of them screenshot a state of your UI and compare it against an approved baseline; they differ in where the screenshot is rendered, where the baselines live, how a change gets reviewed, and how they bill. For which one to pick in your situation, see the best visual regression testing tools in 2026.

A visual regression testing tool renders each state of your UI, takes a screenshot, and compares it against a baseline you have approved. When a pixel changes that you did not intend, it fails the pull request so the regression is caught before it merges. The tools differ in where the baselines live, how a change is reviewed, and whether anything helps you decide if the change was intended.

Compare them on five things: where the screenshot is rendered, where the baselines live, how a change is reviewed, what they can capture, and how they bill. A screenshot taken on your own CI machine shifts with its OS, browser version, and fonts, while a vendor's fixed render environment keeps that noise out of the diff. Baselines either sit in your repo as image files or are hosted per branch. A change reaches you as a failed test, a review queue, or a queue an AI or your coding agent has already been through. Capture ranges from Storybook stories to end-to-end tests to native apps, and most hosted tools bill per rendered screenshot.

• Per-story or per-page screenshot capture in a real browser, diffed against a baseline.
• A way to review a change and accept or reject it.
• A way to skip unchanged work, in most of them, so a small PR does not re-render everything.
• Cross-browser rendering, to varying degrees.

Chromatic is best for Storybook teams that want axe accessibility checks in the same run (on every plan, each billed as an extra snapshot) and Microsoft Edge on its paid plans; it is the Storybook team's own visual testing service. UI Verify matches the published, permalinked Storybook - it re-hosts a live, browsable one at a stable per-branch URL - and adds agent review over MCP, an AI judge, and collaborative review where designers and engineers comment on the exact pixels and approve or reject, over a bigger free tier. See UI Verify vs Chromatic.

Percy is best for teams that already run BrowserStack or need native mobile apps and real devices; it is BrowserStack's visual testing product, with an AI Visual Review Agent on its paid plans. Against UI Verify, the differences are pricing, an archive-replay setup that captures every test with no snapshot calls, and a standalone service rather than a module inside BrowserStack. See UI Verify vs Percy.

Argos is best for teams whose CI already takes screenshots, including from Puppeteer or WebdriverIO, or who want open-source code. You capture with its SDKs, which stabilize the page first, upload the screenshots, and Argos diffs them, hosts the baselines, and lets a coding agent review and approve a build over MCP. UI Verify renders the screenshots for you by replaying your existing tests, and can add an AI judge on each change. See UI Verify vs Argos.

Applitools is best for a large organization that wants one enterprise platform: perceptual Visual AI with tunable match levels, a cross-browser and device grid, test generation, and accessibility tooling. UI Verify is lighter and self-serve, with an AI judge aimed at the pull request decision. See UI Verify vs Applitools.

Playwright's built-in toHaveScreenshot is enough for a small suite you review yourself: it is free and local, diffing against baseline images committed to your repo. UI Verify moves the baselines, review, and judgment into a hosted service with a PR check. See UI Verify vs Playwright screenshots.

UI Verify is best for teams that write pull requests with coding agents: the agent checks what its change did to the UI while it works, before it opens a PR, and every PR is checked again. Over MCP the agent reads each change as a cropped before-and-after, with the AI judge's verdict when the judge is on, and accepts the intended baselines, while automatic flake detection and inherited-change labels keep noise out of the queue. It captures your existing Storybook, Playwright, and Vitest tests by archive-replay and includes 10,000 snapshots a month free (50,000 for open-source projects).

Most hosted tools here bill per rendered screenshot (Applitools meters checkpoints, and Playwright's comparison is free), so the cost lever is the same on all of them: render fewer states. On price, UI Verify includes 10,000 snapshots a month free (50,000 for open-source projects), and paid plans start at $89 a month for 30,000 snapshots with $0.004 for each snapshot beyond your plan. Chromatic and Percy include a smaller free allowance and charge more per extra snapshot, while Argos prices its paid plan close to UI Verify's and charges less for a Storybook screenshot past the plan (pricing pages in the table's sources). To cut the bill on any of them, reduce your snapshot count and skip the stories a pull request does not touch.

It depends on what you are optimizing for. Chromatic if you want axe accessibility and Microsoft Edge in the same run. Percy if you need a device cloud or already run BrowserStack. Argos if your CI already captures screenshots, including from Puppeteer or WebdriverIO, or you want open-source code. Playwright's built-in comparison if you have a handful of screenshots and want everything free and local. Applitools if you need an enterprise cross-browser grid with test generation. UI Verify if your pull requests come from coding agents and you want the agent to review each visual change over MCP, with an AI judge, pixel-anchored comments, and a bigger free tier over the tests you already have.

No. Every tool here can capture real pages from end-to-end tests, not only components. UI Verify captures Storybook stories, Vitest browser-mode components, and real pages driven by Playwright through archive-replay, so you can visually test a component library, a Next.js app, or a staging URL. See the Storybook, Playwright, and Vitest quickstarts.