We Tested 7 Chrome Ad Blockers Under Manifest V3 (2026)

Key takeaways

• uBlock Origin Lite neutralized 14 of 14 test scripts, the only clean sweep. SuperchargePerformance and Ghostery followed at 10 of 14.
• Adblock Plus and AdBlock (96M combined users) blocked the fewest: 4 of 14. Both ship an 8,257-rule Acceptable Ads allowlist, enabled by default, that explicitly permits Google’s gpt.js ad tag.
• Defaults beat filter size. A fresh Ghostery install blocks nothing until you click consent. AdGuard’s tracking filter ships disabled. Package rule counts predicted almost nothing.

On August 7, 2026 we loaded seven Manifest V3 ad blockers into Chromium 148, one per isolated browser profile, and measured what each one stops. uBlock Origin Lite neutralized all 14 live test scripts. Adblock Plus and AdBlock blocked 4 — both ship Google’s gpt.js ad tag on an Acceptable Ads allowlist that is enabled by default.

Every extension was the current Chrome Web Store build, installed fresh with settings untouched, because that is how most of the internet runs an ad blocker: install, close the tab, never open the options page.

The Results, Ranked

Fourteen real ad and analytics scripts (AdSense, Google Publisher Tag, Criteo, Taboola, Outbrain, Amazon’s apstag, IAS, Rubicon/Magnite, Google Analytics, Meta Pixel, Hotjar, Scorecard, Chartbeat, LinkedIn Insight) injected as <script> elements from a live page, exactly the way publishers load them. Lower “loaded” is better.

*Ghostery’s misses come with an asterisk: the AdSense and Google Analytics requests matched Ghostery rules in Chrome’s DNR test API, yet the scripts loaded. That behavior is consistent with redirect-to-stub handling rather than a plain miss. We scored strictly: if the real script executed, it counts as through.

The user-count column deserves a second look. The two most-installed ad blockers on Chrome finished last. The winner is tied for third-largest, level with AdGuard at 16M. Install base measures marketing reach and inertia, not blocking.

How We Ran the Test

Test note for the record: all measurements taken August 7, 2026, on Chromium 148 (headless, CDP-driven), each extension loaded unpacked from its current Chrome Web Store package into its own throwaway profile. Three instruments, cross-checked:

1. Manifest audit. We unpacked each CRX and counted every declarativeNetRequest ruleset: which ship, which are enabled, how many rules each contains. This is the extension’s blocking capacity as Chrome sees it, not as the marketing page describes it.
2. Chrome’s own rule engine. declarativeNetRequest.testMatchOutcome against 10 ad-serving URLs per extension, asking Chrome directly: would you block this request? This is what caught the Acceptable Ads allow rules; a request can “match” a rule whose action is allow.
3. Live page probe. From a real page, each browser injected the 20 script URLs and recorded load, block, or timeout. A no-extension baseline loaded the same set first; 6 endpoints failed even unblocked (dead test IDs), leaving 14 valid scripts. Every extension faced the identical 14.

The combination matters. Instrument 2 told us why something loaded (allow rule vs. no rule). Instrument 3 told us what actually happens on a page. Where they disagreed (uBOL requests that “matched” but still loaded), the rule files themselves settled it: those are redirect actions pointing at neutered local stubs, gorhill’s long-standing technique for breaking trackers without breaking the sites that expect them.

One scope caveat: this test measures network-level blocking of ad and tracker scripts. It does not measure cosmetic filtering, YouTube video ads, or popups. Those are different mechanisms and a different test. Our earlier five-blocker feature comparison covers that ground.

The Allowlist Nobody Mentions at Install

The most consequential finding sits in Adblock Plus and AdBlock, and it is identical in both because both belong to eyeo GmbH. Their shipped packages contain the same eyeo ruleset files — same UUIDs, same rule IDs — and three arrive enabled: EasyList (57,191 rules), an anti-circumvention list (1,661), and the Acceptable Ads exception list: 8,257 allow rules.

When our probe requested securepubads.g.doubleclick.net/tag/js/gpt.js, the tag that loads Google display ads on millions of pages, Chrome’s rule engine reported a match: allow rule #64135, filter ||g.doubleclick.net^*/gpt.js. Criteo’s publisher tag matched allow rule #58207. Both scripts then loaded and executed in the live test. That is not a gap in coverage. It is a business model: publishers and networks pay eyeo for placement on the exception list, and the default user never learns the toggle exists.

To be fair where fairness is due: the setting is disclosed, documented, and takes one checkbox to turn off, and with Acceptable Ads disabled both extensions enforce the same EasyList core that caught apstag, Rubicon, and Meta Pixel in our test. But we tested defaults, and the defaults wave Google’s ad infrastructure through.

Defaults Decide More Than Filters Do

Three more extensions proved the same rule from different angles.

Ghostery ships every ruleset switched off. All 33 rulesets, 90,435 rules, disabled in the manifest until you click “Enable Ghostery” on its consent screen. Fresh install: 0 of 20 endpoints blocked. After consent: 10 of 14. Ghostery does this for GDPR reasons and asks clearly at install. A defensible design, but it means a skipped welcome tab leaves you unprotected while the toolbar icon sits there looking helpful.

AdGuard enables its Base filter and nothing else. One ruleset on (77,214 rules) of the 51 shipped (329,385 total). The Base filter swept the classic ad-tech domains — AdSense, gpt.js, Outbrain, apstag, IAS, Rubicon — and missed every pure analytics endpoint we threw: Google Analytics, Meta Pixel, Hotjar, Scorecard, Chartbeat, LinkedIn. Those live in AdGuard’s Tracking Protection filter, which ships disabled. Turn it on and this picture should change substantially; out of the box, AdGuard is an ad blocker, not a tracker blocker. Worth noting for anyone comparing listings: the main AdGuard listing (16M users) and the separate “AdGuard AdBlocker (Beta)” listing shipped the identical 5.4.3.119 build when we pulled both.

Stands runs lean and it shows both ways. The whole package is 15 MB, thirty-five times smaller than Adblock Plus, and its 15,416 rules, all enabled, outblocked ABP’s default configuration 6 to 4, including Google Analytics, which ABP allowed through. It still missed the native-ad networks (Criteo, Taboola, Outbrain) and Meta Pixel.

Rule Counts Are Marketing Numbers

Set the shipped-package sizes against the measured results and the correlation collapses:

The 1.6-million-rule packages finished last; the winner ships 69,869. Most of ABP’s bulk is 30-plus regional and alternate rulesets waiting disabled for users who never enable them. Chrome’s architecture explains why vendors ship this way: Manifest V3 guarantees each extension 30,000 static rules and shares a larger global pool across all installed extensions, so bundling everything and enabling little is free. But it means the number on the listing page describes the warehouse, not the guard at the door.

What the winner does differently is technique, not volume. Seven of uBOL’s fourteen stops were redirect rules serving local stub scripts (a fake analytics.js, a fake Scorecard beacon) so pages that expect those objects keep working while the real payload never arrives. No other extension in this test showed that pattern as consistently.

Where Our Own Extension Landed

SuperchargePerformance is ours, so hold it to the same standard. It stopped 10 of 14: the full prebid/ad-server cluster (gpt.js, Criteo, Taboola, Outbrain, apstag, IAS, Rubicon), plus Google Analytics, Meta Pixel, and Chartbeat. It let four through: the AdSense loader (adsbygoogle.js), Hotjar, Scorecard, and LinkedIn Insight. That places it joint second on network blocking with 195,058 of its 195,088 rules enabled by default, no paywall, and behind uBlock Origin Lite, which is the honest headline.

The blocklist is one layer of a performance extension rather than the whole product; the same install suspends idle tabs and blocks YouTube and Twitch video ads through separate, platform-specific machinery that a network probe like this one does not exercise. No other extension in this table touches video ads at all. If your sole requirement is maximum general-purpose script blocking, this test says install uBOL. If ads are one of several things you want off your CPU, the trade reads differently.

Which One Should You Install

The conditional version, from the data:

• You want one blocker and maximum coverage: uBlock Origin Lite. 14 of 14, and the stub-redirect technique breaks trackers without breaking pages.
• You run AdBlock or Adblock Plus today: open settings and untick Acceptable Ads. That single checkbox is the difference between shipping Google’s ad tag and blocking it.
• You installed Ghostery at some point: open its panel and confirm it is actually enabled. If you never clicked the consent screen, it has been decorative.
• You use AdGuard and care about analytics trackers: enable the Tracking Protection filter; the default config allowed every analytics endpoint we tested.
• You also want tab suspension, video-ad handling, and a RAM readout in the same install: SuperchargePerformance, with the measured caveat that its general blocking trails uBOL by four scripts.
• Manifest V2 nostalgia: the browser side ended with Chrome 138 in July 2025 and the last enterprise remnants left in Chrome 151 on July 28, 2026. The August 31, 2026 deadline everyone cites only deletes the remaining store listings. Nothing you run today changes that weekend — every extension in this test is already MV3.