Free tool Gzip compression test: is your page compressed?
Enter a URL. The test asks the server for Brotli, gzip and no compression in turn, and shows which encoding it returns each time, how many bytes travel over the network, how much that saves and whether caches are told the response varies by encoding. If the server does not report the compressed size, your browser estimates the gzip size itself.
- br, gzip, zstd, identity
- Size and savings
- Sent only when you click
Test compression for a URL
Test an HTML page, a CSS or JavaScript file, an SVG or a JSON endpoint. Images and fonts in modern formats are already compressed. Until you click Check, the result is a sample for wikipedia.org.
check.tapcub.com, which fetches it from Cloudflare’s network with the user-agent shown and a From: tools@tapcub.com header, then returns the status codes, response headers and at most 512 KB of the page. One test uses three requests (four with zstd). The gzip estimate is computed in your browser and sends nothing. The service records only the host name, status and timing for rate limiting, not the path, query or page content, and stores no results. Limits: 10 requests a minute and 200 a day per IP address.A page or a file. https:// is added if you leave it out.
Uncompressed size (bytes)
122,506
The decoded body.
gzip on the wire (bytes)
29,183
From the server’s Content-Length.
gzip estimate (bytes)
29,243
Recompressed in your browser at the default level.
Brotli on the wire (bytes)
—
From the server’s Content-Length.
Saving
76.2%
Smallest reported (or estimated) size versus uncompressed.
Checks
- OKgzip is enabled.
- NoteBrotli is not offered: asking for br only returns the file uncompressed. Browsers then use gzip, which is fine; Brotli usually saves a bit more on text.
- WarningThe gzip response has no Vary: Accept-Encoding. A shared cache could hand the compressed copy to a client that did not ask for it.
Per request
Each row asks for one encoding. “Not reported” means the server streamed the body without Content-Length.
| Accept-Encoding sent | Content-Encoding returned | Transfer size | Uncompressed | Vary |
|---|---|---|---|---|
br | none | 122,506 B | 122,506 B | — |
gzip | gzip | 29,183 B | 122,506 B | — |
identity | none | 122,506 B | 122,506 B | — |
Share these inputs: the values go after the # in the link, so they never reach a server.
https://tapcub.com/tools/compression-test
Version v1.0Updated Update history
What the numbers mean
Compression only changes how many bytes travel. The browser decodes the same page either way.
Content-Encoding
The encoding the server actually used. gzip and br (Brotli) are what browsers accept everywhere; zstd is accepted by recent Chrome and Firefox. “none” means the body was sent as is.
Transfer size
Bytes on the wire, taken from Content-Length. Many servers stream (chunked) and send no length, so the size is not visible from outside; that is when the browser estimate helps.
gzip estimate
Your browser recompresses the returned text with CompressionStream at the default level. Servers often use the same level, so it lands close: within 0.2% for the wikipedia.org sample. Your browser cannot produce Brotli, so there is no Brotli estimate.
Vary: Accept-Encoding
Tells caches (CDN, proxies) to keep separate copies for clients that do and do not accept compression. Without it, a cache can serve gzip to a client that cannot read it.
Lighthouse’s “Enable text compression” audit gzips every uncompressed text response and flags those of at least 1.4 KiB where compression would save 10% or more. If a large file shows no saving here, start there.
Three steps
Nothing is requested until you click Check.
-
1
Enter a URL
Your homepage first, then the largest CSS and JavaScript files from the browser’s Network tab.
-
2
Read the checks
gzip should be on for every text response. Brotli is a bonus. Vary should be set wherever compression is.
-
3
Turn it on where missing
In Nginx it is
gzip on;plusgzip_types; Apache uses mod_deflate; most CDNs compress text by default. Test again afterwards.
How the test works
check.tapcub.com fetches the URL from Cloudflare’s network once per encoding:
- Three GET requests with
Accept-Encoding: br,gzipandidentity(pluszstdif ticked); redirects are followed - Content-Encoding, Content-Length and Vary are read from each final response. Content-Length on a compressed response is the compressed size (RFC 9110 §8.6)
- The service decodes the body before returning it, so it cannot measure compressed size itself; when Content-Length is missing the size shows as “not reported”
- Saving = 1 − smallest compressed size ÷ uncompressed size; if no size is reported, the gzip estimate from your browser is used
| Type | Compress? | Why |
|---|---|---|
| HTML, CSS, JavaScript, JSON, XML, SVG | Yes | Plain text with lots of repetition |
| TTF / OTF fonts, ICO | Yes | Uncompressed binary formats |
| WOFF2 fonts | No | Already Brotli-compressed inside |
| JPEG, PNG, WebP, AVIF, GIF | No | Already compressed; gzip may even add bytes |
| MP4, WebM, ZIP, PDF (mostly) | No | Already compressed |
wikipedia.org: gzip yes, Brotli no
The sample on this page was fetched on 2026-11-17.
| Asked for | Got | Transfer size | Reading |
|---|---|---|---|
br | none | 122,506 B | Brotli not offered, so the server sent the page uncompressed |
gzip | gzip | 29,183 B | 76.2% smaller than 122,506 B |
identity | none | 122,506 B | Correct: no compression when none is accepted |
| Browser estimate | gzip level 6 | 29,243 B | 60 bytes off the real gzip size |
Browsers send gzip, deflate, br (and zstd in recent versions), so real visitors get the gzip version here. The responses carried no Vary header, which the test flags as a warning.
Why compression is off when you think it is on
Compression is usually configured once and never looked at again.
Only text/html compressed
Nginx compresses only text/html unless gzip_types lists CSS, JavaScript, JSON and SVG too.
Compressing twice
The app compresses and the proxy compresses again, or images get gzipped. Compress text once, at the edge or the server.
Origin fine, CDN not
Some CDN plans or cache rules strip compression. Test the public URL, not the origin server.
Missing Vary
A cache without Vary: Accept-Encoding can serve compressed bytes to a client that cannot decode them.
See speed as real visitors get it
Compression is one input. TapCub measures what visitors experience page by page, so you can see whether a change helped.
Web Vitals per page
Real-user LCP, INP and CLS for each page, to see whether a script (ours included) slows it down.
Learn moreErrors next to speed
JavaScript, resource and API errors with affected sessions, in the same report.
Learn moreBots counted separately
Five layers and 197+ rules keep crawlers out of the denominator by default.
Learn moreGzip compression test questions
Still have a question?
Chat with the team behind TapCub — we usually reply within a few hours on business days.
How do I check if gzip is enabled?
Enter the URL above. The gzip row shows gzip under “Content-Encoding returned” when it is on. In your browser, open developer tools, Network tab, click the request and look for content-encoding in the response headers.
Is Brotli better than gzip?
For text, Brotli at typical settings usually produces smaller files than gzip, and every current browser accepts it over https. Gzip is still fine; enabling Brotli is an extra, not a fix.
Why is the compressed size “not reported”?
The server sent the body in chunks without a Content-Length header. Our service receives the body already decoded, so the compressed byte count is lost. The browser estimate fills the gap for gzip.
How accurate is the gzip estimate?
It recompresses the same text at the default gzip level, which most servers also use, so it is usually within a few percent. Servers set to a different level, or pages over 512 KB (only the first 512 KB is estimated), can differ more.
Should images be gzipped?
No. JPEG, PNG, WebP, AVIF and WOFF2 are already compressed; gzip adds CPU time and can make them slightly larger. SVG is text and should be compressed.
What does Vary: Accept-Encoding do?
It tells caches that the response depends on the Accept-Encoding request header, so they store the compressed and uncompressed versions separately. Most servers add it automatically when they compress.
Is zstd worth enabling?
It is optional. Chrome (from version 123) and Firefox (from 126) accept it; others fall back to Brotli or gzip. Turn it on if your CDN or server supports it, but keep gzip and Brotli.
Update history
New features, fixes and data refreshes for this tool, newest first.
-
· v1.0
LaunchedFirst release.
Data this tool relies on
- Sample fetches (github.com, apple.com, wikipedia.org, example.com, bbc.com) — checked 2026-11-17, reviewed every 6 months (next review by 2027-05-17)
Missing something or found an error? Tell us — requests that come up often go into the next update. All tool updates
Check the rest of the page weight
Compression shrinks text. Unused code and heavy scripts are the next place to look.