Illustration of a website product page in Labvanced with a green outlined tracked area, gaze heatmap spots, a cursor, and scroll icons, beside the title Website Eye Tracking and Behavior Tracking

Website Eye Tracking & Behavior Tracking

Labvanced Web Bridge is a free Chrome extension for website eye tracking and behavior tracking. It embeds a live website in a Labvanced study and records where participants look, click, and scroll, using Labvanced's webcam eye tracking, which was validated against an EyeLink 1000 in a peer-reviewed study (Kaduk et al., 2024).

What Is Labvanced Web Bridge?
What Labvanced Web Bridge Can Do
How Web Bridge Differs
Research You Can Do With Labvanced Web Bridge
Data Collected
Benchmarking Accuracy
How Labvanced Web Bridge Tracks a Real Website
How Website Tracking Data Behaves
Setting Up Labvanced Web Bridge
Limitations
Frequently Asked Questions

What Is Labvanced Web Bridge?

Labvanced Web Bridge is a free Chrome extension that connects Labvanced to real, external websites. Normally, a research platform can only show participants stimuli built inside it, such as images, videos, forms, and custom pages. Web Bridge changes that: it lets a researcher bring an actual, live website, such as a product page, a news site, or a web app, directly into a study, and then measure how participants behave on it.

It's especially useful for controlled UX and behavioral research, where seeing how someone reacts to the real thing matters more than a recreation of it. It isn't limited to that.

What Labvanced Web Bridge Can Do

Web Bridge is part of Labvanced's full experiment platform, so the same study can randomize which version of a page a participant sees, combine website tracking with Labvanced's other measures, and export everything in one consistent format.

Real Websites, Not Mockups

Participants interact with your actual website, live, inside the study. What gets tracked is the real page your users will actually see.

Everything In One Place

Gaze, clicks, hovers, and scrolling are all captured on the very same tracked element, so every signal lines up without extra work.

Tracks Elements, Not Screen Positions

Mark something once. Tracking keeps following it correctly through scrolling, resizing, and even pop-ups or menus that only appear after a click.

How Web Bridge Differs

Other ways to test a real website differ in what they need from the site and what they can record on it.

The question	Other approaches	Web Bridge
Does the website need changes?	RealEye, a webcam eye-tracking service, needs its JavaScript SDK on the site and a `data-re-aoi-name` attribute on every tracked element for its embedded website test. Microsoft Clarity, a heatmap and session-recording tool, needs its tracking code installed on the site.	No. You paste the URL of the website you want to test and mark elements by clicking them in the Labvanced editor.
Can it record what happens inside an embedded page?	A plain embed of another website, such as the Labvanced `I-Frame` object, cannot be read by the page around it, so clicks and scrolling inside it are not captured. Microsoft Clarity also cannot capture third-party iframes.	Yes. The extension reads clicks, hovers, mouse movement, and scrolling from inside the embedded page.
Does it keep up with scrolling and changing pages?	RealEye's embedded website test analyzes a screenshot of the site and does not support horizontal scrolling.	Yes. Marked elements are followed through scrolling, resizing, animation, and content that loads after the page opens.
Does the data come from real participants?	Predictive tools such as Attention Insight estimate attention with an AI model instead of recording real participants.	Yes. It records where real participants look and click.
Does it need special hardware?	Pupil Labs, an eye-tracker maker, offers a Web AOI tool that requires its Neon eye tracker.	No. It uses the participant's webcam.

For what an AI-predicted heatmap can and can't tell you compared with recording real participants, see AI Can Predict Where You'll Look. It Still Doesn't Know What You'll Do.

Last reviewed: 2026-09-25. Details were checked against RealEye's Knowledge Base, Microsoft Clarity's setup documentation, Attention Insight's knowledge base, and Pupil Labs' Web AOI documentation. The browser restriction on reading embedded pages is described in MDN's same-origin policy documentation.

Research You Can Do With Labvanced Web Bridge

Mock running shoe product page with a gaze heatmap over the shoe photo, product name, price, color swatches, and Add to Cart button

E-Commerce

Embed a real product page. See which photos people actually look at, which buttons get clicked, and how far they scroll before adding something to their cart.

Mock news page with a gaze heatmap concentrated on the article headline and the mountain lead photo beside it

News & Media

Embed a live news article. Find out which headlines get noticed first, and how attention changes as an article gets longer.

Mock website navigation bar with the Solutions dropdown open and a dashed path tracing movement from the Solutions menu to Pricing

Navigation Testing

Embed a real menu or multi-page flow. See whether people can actually find what they're looking for, and where they hesitate.

Two mock bedding landing pages labeled A and B, with different headlines and Shop Now button colors, each captioned with the share of visitors who clicked, 42 percent and 67 percent, and a progress bar

A/B Testing

Embed two versions of the same page, randomly assign participants to each with Labvanced's built-in counterbalancing, and compare the results directly.

Mock children's educational quiz page asking Which animal is a lion, with a lion illustration and three answer cards (lion, elephant, zebra), and a gaze heatmap strongest on the lion illustration and the lion answer card

Developmental Psychology

Embed a real children's educational website and track how young participants navigate it. This is useful for research on digital literacy and how attention develops online.

Mock news article page with the headline Can We Trust the News in a Polarized World, a photo of people outside a building, and a gaze heatmap over the headline, the photo, and the author byline

Media & Political Psychology

Embed a real news article and track whether people's eyes ever reach the byline or source, before they decide whether to trust it.

Data Collected

Embedding a real website turns ordinary browsing into research data. With Web Bridge, you can find out:

Where visitors look first, and for how long
Which elements never get looked at, or get skipped past entirely
How attention moves from one part of a page to another before someone clicks
Whether what people click matches what they looked at first, or attention and action pull in different directions
How long people linger on one area, which can mean they're reading closely or are stuck looking for something

Pick anything on the page, such as a product photo, a headline, or a button. Here is the data recorded for it:

You get	What it tells you
Gaze	Where people actually look, and for how long
Element position	The exact location and size of what you're tracking, updated live as the page scrolls
Clicks	What people click, and exactly where
Hovers	What catches attention even without a click
Mouse movement	How the cursor moves across the page
Scrolling	How far people scroll before they act

Every result exports cleanly to CSV, XLSX, or TSV, ready to analyze.

Benchmarking Accuracy

Web Bridge combines two of Labvanced's core measurement engines. Gaze data comes from Labvanced's webcam eye tracking, tested against a professional eye tracker (the EyeLink 1000) in a peer-reviewed study: an overall accuracy of 1.4° and a precision of 1.1°, improving to 1.3° accuracy and 0.9° precision at the center of the screen (Kaduk et al., 2024).

Click, hover, and cursor data come from Labvanced's mouse tracking and are recorded on the same tracked element as the gaze data.

How Labvanced Web Bridge Tracks a Real Website

Web Bridge enables controlled behavioral research on live, third-party websites: participants interact with the real page while Labvanced records their eye movements, clicks, hovers, and scrolling.

Making that possible means working around two separate browser restrictions. First, many websites set a security header, X-Frame-Options or Content-Security-Policy, specifically to stop themselves from being embedded inside another page at all. Web Bridge lifts that restriction only in browser tabs where Labvanced is open. Other tabs are unaffected, and the change is removed when the tab closes or leaves Labvanced. Second, even once a site is embedded, browsers don't normally let mouse activity inside that embedded page reach the page around it. This is a separate restriction. Web Bridge's extension handles that too, reading click, hover, and movement events directly from inside the embedded page and relaying them back to the study in real time.

To use it, researchers add a Website Frame to a study and point it at a real URL. From there, marking what to track is a simple click-and-select action: click on a product photo, a headline, or a button, and Labvanced locks onto that exact element, not its position on screen. Labvanced continuously monitors the page, so tracking automatically adjusts for scrolling, for elements that only appear after a click, like a pop-up or a dropdown, and even for animations or elements that resize or move on their own. Every tracked area is a simple rectangle, and nothing is tracked automatically: a researcher chooses exactly what matters before a study runs.

Property	Specifications
Browser	Google Chrome
Built On	Labvanced's peer-reviewed webcam eye tracking (Kaduk et al., 2024) and its mouse tracking
What It Tracks	Gaze, clicks, hovers, and scrolling on any element you choose
Privacy	Only runs inside websites that Labvanced embeds. Text in password and payment fields is never read

How Website Tracking Data Behaves

For designing a study around this data, a few behaviors are worth knowing up front.

Coordinate spaces differ by interaction type. Click, mouse-down, mouse-up, and mouse-move events report page-relative coordinates, which is the position across the embedded page's full scrollable area, including whatever has already been scrolled past. Hover and leave events instead report element-relative coordinates, which is the position within the tracked element itself. Both use the same coordinate labels in the interface, so decide up front which interaction type a study needs. Combining the two in one exported column mixes two different coordinate systems with no warning. Gaze data recorded on a tracked element uses the same page-relative space as clicks, so gaze and click data from the same element line up directly without any conversion.

Tracking updates continuously, not only on scroll. Every tracked element's position and size is re-measured on every frame, which is what lets tracking correctly follow animations, CSS transitions, and elements that resize dynamically, not just scrolling or repositioning.

Scrollable regions inside a tracked element get their own coordinate space. If a tracked element has its own internal scrolling, such as a chat panel or a scrollable product list, marking it as a scrollable container exposes that region's own scroll position and content size separately from the page's overall scroll, so nested scrolling isn't conflated with the outer page.

Mouse events offer three targeting modes. A mouse event can fire for one specific tracked element, for any tracked element without picking one individually, or for anywhere on the embedded page regardless of whether it's a tracked element at all. This is useful for capturing broad interaction patterns without marking every element in advance.

Setting Up Labvanced Web Bridge

Web Bridge is a free Chrome extension. Both you and your participants need it installed. It takes less than a minute, and Labvanced walks participants through it automatically if they don't already have it.

For deeper reference once you're building a study, see Selecting and Adjusting Website AOIs for the full AOI anchoring controls. For building click, hover, and movement events, see Tracking Mouse Behavior on a Website AOI.

Limitations

Web Bridge runs in Google Chrome only, and the researcher and every participant need the free extension installed. Using a Website Frame in a study requires the Ultimate license. Some sites cannot be embedded: amazon.com and ebay.com are blocked in the editor because of their bot protection, and other sites that detect embedding may not work. Web Bridge tracks the elements a researcher marks rather than everything on a page.

Frequently Asked Questions

Does Web Bridge track people on other websites?

No. Tracking only starts inside a website that Labvanced itself has embedded. On every other page, the extension exits immediately and reads nothing from the page. It only checks whether a browser tab is on labvanced.com, to decide when embedding is allowed.

Do participants need to install anything?

Yes, the same free extension researchers use. If a participant doesn't already have it, a prompt walks them through installing it before the study begins, usually well under a minute.

How does tracking keep working if the page changes?

Every tracked element is locked onto the actual part of the page, not a fixed screen position. Labvanced continuously monitors the page, so if a participant scrolls, or new content appears, like a pop-up or an expanding menu, or an element resizes or animates on its own, tracking follows it automatically.

Does it automatically detect what to track?

No, and that's intentional. A researcher marks exactly what should be tracked, such as a photo, a headline, or a button, by clicking on it directly in the editor. Nothing is guessed or tracked without that manual step.

Can I track clicks and mouse movement without eye tracking?

Yes. Mouse tracking (clicks, hovers, and movement) works independently of eye tracking. You can record one, the other, or both together on the same tracked element.

Is this data as accurate as Labvanced's regular eye tracking?

Yes. It's the exact same engine, validated against an EyeLink 1000 with an overall accuracy of 1.4° and 1.3° at the center of the screen (Kaduk et al., 2024). What's different here is using it on a real external website, not the underlying accuracy.