Is Similarweb - Website Traffic, AI Traffic & SEO Checker safe?
Similarweb is high risk. At startup Similarweb fetches a settings file from an S3 bucket with no signature or format check. It drives feature switches and rules that rewrite addresses before they reach Similarweb. Sessions only returned analytics switches.…
Similarweb findings
Downloads settings from an Amazon S3 bucket at startup, not similarweb.com
At startup Similarweb fetches a settings file from an S3 bucket with no signature or format check.
It drives feature switches and rules that rewrite addresses before they reach Similarweb.
Sessions only returned analytics switches.
What actually happens
You start Chrome, or install the extension, with Similarweb enabled.
No page visit, click, or interaction with the extension is needed.
The extension downloads a settings file from an Amazon S3 bucket and saves it into your synced Chrome storage.
The file it receives decides which analytics systems run and, through a second code path, how the URLs of pages you visit are rewritten before being sent to Similarweb.
Captured request
HTTP 200. Response body captured verbatim: { "mixpanel": { "enabled": true, "config": { "persistence": "localStorage" } }, "matomo": { "enabled": true } }
What's stored on your device
The S3 file is kept as received. It's written to synced storage, not local, so it travels to every Chrome profile you're signed into.
- Location
- chrome.storage.sync, key 'config'
{ "matomo": { "enabled": true }, "mixpanel": { "config": { "persistence": "localStorage" }, "enabled": true }}The code that does this
Startup fetch and persistence, from the extension's shipping background bundle.
The config endpoint constant
background/background.js (deobfuscated)const ENDPOINTS = { CONFIG: "https://sw-extension.s3.amazonaws.com/config.json", DATA: "https://data.similarweb.com/api/v1/data", GLOBAL_RANK: "https://rank.similarweb.com/api/v1/global", IDENTITY: "https://data.similarweb.com/api/v1/identity"};const STORAGE_KEYS = { CONFIG: "config", // ...};The fetch-and-store routine
background/background.js (deobfuscated)async function fetchRemoteConfig() { try { const response = await fetch(`${ENDPOINTS.CONFIG}?v=${Date.now()}`); const config = await response.json(); await storageSet(STORAGE_KEYS.CONFIG, config); // chrome.storage.sync.set({ config }) } catch (err) { logError(err); }}Called from the top-level startup routine
background/background.js (deobfuscated)(async function startup() { try { initA(); initB(); installIdentityHeaderListener(); await fetchRemoteConfig(); } catch (err) { logError(err); }})();The code that does this
How the stored configuration is consumed: settings loader and URL-rewriting rule builder.
Settings applied from a response header on the data endpoint
background/background.js (deobfuscated)fetch(this.dataEndpoint, { method, headers, body }).then(response => { const sessionHeader = response.headers.get("x-session-id"); if (sessionHeader) { try { const settings = JSON.parse(atob(sessionHeader)); runHook.setSettings(settings); // replaces the live settings object } catch (e) {} } // ...});The rule builder that consumes config.data
background/background.js (deobfuscated)createGetClasses(pii) { if (typeof pii === "undefined") return; // config.data is either a base64-encoded JSON string or a plain object const rules = typeof pii.data === "string" ? JSON.parse(atob(pii.data || "e30=")) : pii.data; // "urlparams" entries are "keyRegex=valueRegex[=type]" triples const urlparams_m = {}; Object.keys(rules.urlparams || {}).forEach(name => { const parts = rules.urlparams[name].split("="); if (parts.length === 2 || parts.length === 3) { urlparams_m[name] = { kr: parts[0], vr: parts[1], t: parts[2] }; } }); rules.urlparams_m = urlparams_m; const handlers = []; try { if (typeof rules.sitewhitelist === "string") handlers.push(new SiteWhitelistHandler(true, rules.sitewhitelist)); if (typeof rules.blacklist === "object" && Object.keys(rules.blacklist).length) handlers.push(new BlacklistHandler(true, rules.blacklist)); if (typeof rules.paramwhitelist === "string" && typeof rules.urlparams_m === "object" && typeof rules.paths === "object" && Object.keys(rules.paths).length) handlers.push(new ParamWhitelistHandler(false, rules.paramwhitelist, rules.urlparams_m, rules.paths)); } catch (e) {} this.holdings_list = handlers; // applied to URLs before they are sent}Hosts hardcoded in the background bundle's endpoint table.
- sw-extension.s3.amazonaws.com
S3 bucket serving config.json, fetched every worker start and applied with no integrity checks. Not similarweb.com; the bucket's own access control governs what the extension gets.
- data.similarweb.com
Similarweb's data/identity API. Its x-session-id header decodes to base64 JSON fed into the same settings loader, a second server-controlled config channel.
- rank.similarweb.com
Similarweb's global rank API, queried for site metrics after URLs have been processed by the configured rewriting rules.
What we observed, and what we did not
The startup fetch and the storage write are confirmed: two independent dynamic-analysis sessions each captured the background service worker performing GET https://sw-extension.s3.amazonaws.com/config.json, and the post-session storage snapshot held the exact response body under chrome.storage.sync['config'].
The configuration served at the time of testing contained only Mixpanel and Matomo feature toggles. The urlparams, sitewhitelist, blacklist, paramwhitelist and paths fields that the rule builder is written to consume were not present in that response, so we did not observe URL-rewriting rules being delivered. Those code paths are present in the shipping bundle and are reached from the same stored config object; what they do on a given day depends on the document the bucket returns at that moment.
The vendor's server can make Similarweb read text on any site
Code analysis shows a remote-config engine runs on all sites, including AI chat sites, and can read page text and input fields as the server directs.
No chat text was observed leaving.
What actually happens
The vendor's config server sends extraction rules.
The extension reads page text and input fields that the rules select on any site.
This includes AI chat sites if a rule targets them.
What the engine can read
- Visible page textelement.innerText
Text of any element the rule selects.
- Input valuesinput.value
What is typed in fields the rule selects.
- Attributes and markupgetAttribute / innerHTML
Attributes and HTML of selected elements.
- XPath queriesdocument.evaluate(...)
Rules can address any node in the page.
The code that does this
Config endpoint and consent text
Readable form
const cfg = await fetch('https://rank.similarweb.com/content/config', {method:'POST', body: JSON.stringify({sid})});// 56 encoded {"e":"<cipher>"} rule objects, decoded in the extensionDestination
- rank.similarweb.com
Serves the extraction rules; also the vendor's data collection host.
Observation
Static analysis finding. This behaviour was identified by reading the shipped extension code and has not yet been reproduced in a live run. The trigger conditions and the exact data sent are read from the code, not from an observed capture.
Sends every page you visit and its referrer to Similarweb
Code analysis shows the extension posts each page address and its referrer to rank.similarweb.com on every page navigation.
What actually happens
You load any web page.
The extension sends the page address and referrer to rank.similarweb.com.
The report is a form-encoded POST to /api/v1/global.
Fields in each report
- Current pagehttps://example.com/account/orders
The address of the page you are viewing.
- Previous pagehttps://example.com/search?q=shoes
The page you came from.
- Referrer headerhttps://www.google.com/
The Referer value sent with the request (illustrative).
- Canonical addresshttps://example.com/account/orders
The canonical link read from the page.
- Tab id1482
Which browser tab produced the visit.
The code that does this
Endpoint and body construction
Readable form
const ENDPOINT = 'https://rank.similarweb.com' + '/api/v1/global';body = 'e=' + encodeURIComponent(data) + '&decode=0';Destination
- rank.similarweb.com
Receives URL, referrer and tab data for each navigation.
Observation
Static analysis finding. This behaviour was identified by reading the shipped extension code and has not yet been reproduced in a live run. The trigger conditions and the exact data sent are read from the code, not from an observed capture.
+3 more findings not shown
If you have Similarweb installed
- Open
chrome://extensionsand turn on Developer mode to show IDs. Find the one with IDhoklmmgfnpapgjgcpechhaamimifchmpand choose Remove. - Block the servers only it was seen contacting:
ca.knock-ai.com,d2c7xlmseob604.cloudfront.net,gtm.similarweb.com,analytics.getlimy.ai,bzr.openai.com,bzrcdn.openai.com,js.knock-ai.com. - Scan your browser to check the rest of your extensions.
Similarweb version history
6.12.17
Mar 4, 2026
+4 more
6.12.22
Aug 3, 2026
6.12.24
Sep 30, 2026
6.12.26
Oct 19, 2026
Where Similarweb sends data
- pagead2.googlesyndication.com/ccm/collect22 calls
From a script it injects into websites · 7 other extensions
- www.google.com/ccm/collect22 calls
From a script it injects into websites · 108 other extensions
- www.google.com/rmkt/collect/…18 calls
From a script it injects into websites · 108 other extensions
- mpps.similarweb.com/track15 calls
From a script it injects into websites · 1 other extension
- rank.similarweb.com/content/config12 calls
From its background script · 1 other extension
- sw-extension.s3.amazonaws.com/config.json10 calls
From its background script · 1 other extension
- ca.knock-ai.com/client/an/…9 calls
From a script it injects into websites · only this extension
- api.mixpanel.com/track8 calls
From its background script · 22 other extensions
- d2c7xlmseob604.cloudfront.net/configuration/…8 calls
From a script it injects into websites · only this extension
- gtm.similarweb.com/g/collect8 calls
From a script it injects into websites · only this extension
- www.google.com/async/hpba5 calls
From a script it injects into websites · 108 other extensions
- analytics.getlimy.ai/e4 calls
From a script it injects into websites · only this extension
- api-js.mixpanel.com/track4 calls
From a script it injects into websites · 17 other extensions
- bzr.openai.com/v1/sdk/…4 calls
From a script it injects into websites · only this extension
- bzrcdn.openai.com/pixel-config/v1/…4 calls
From a script it injects into websites · only this extension
- consentcdn.cookiebot.com/consentconfig/…4 calls
From a script it injects into websites · 1 other extension
- js.knock-ai.com/prod/v/…4 calls
From a script it injects into websites · only this extension
- m.media-amazon.com/images/I/…4 calls
From a script it injects into websites · 33 other extensions
- www.facebook.com/ajax/bz4 calls
From a script it injects into websites · 28 other extensions
- www.google-analytics.com/j/collect4 calls
From a script it injects into websites · 238 other extensions
- chatgpt.com/backend-api/sentinel/…3 calls
From a script it injects into websites · 28 other extensions
- ogads-pa.clients6.google.com/$rpc/…3 calls
From a script it injects into websites · 27 other extensions
- play.google.com/log3 calls
From a script it injects into websites · 38 other extensions
- www.amazon.com/dram/…3 calls
From a script it injects into websites · 55 other extensions
- www.amazon.com/portal-migration/hz/…3 calls
From a script it injects into websites · 55 other extensions
Named inside its requests but not contacted (servers it tested or reported on): 49625052041-kgt0hghf445lmcmhijv46b715m2mpbct.apps.googleusercontent.com, ep.tt, sst.us, www.similarweb.com.
Chrome Web Store listing
- Versions
- 6.12.26 latest of 8 seen since March 4, 2026
- Store rating
- 4.7 from 3,805 ratings
- Last updated
- May 14, 2026
- Size
- 1.22 MiB
- Privacy policy
- www.similarweb.com
