Most compressors are adaptive: they adjust their state depending on previous inputs. A static compressor wouldn't leak anything-- for HTML, it might have preset

Most compressors are adaptive: they adjust their state depending on previous inputs. A static compressor wouldn't leak anything-- for HTML, it might have preset dictionaries for common tags and attributes (instead of <div>, output a single token), and have a static Huffman encoding tuned to "average" HTML pages. The compression wouldn't be nearly as good as DEFLATE since it has to compress each chunk in isolation, but it would still beat plaintext.