Binary to Text

Start by counting the bits

If the total bit count divides evenly by eight, it is probably byte-aligned ASCII or UTF-8. If it divides by seven but not eight, it may be seven-bit ASCII with no padding, which is common in older telecom and amateur radio contexts. If it divides by neither, characters are missing or extra, and no decoder will produce sensible output until that is fixed.

Why output turns to gibberish partway through

The commonest cause is a single dropped or added bit. Everything before it decodes correctly and everything after is shifted, producing text that starts fine and degenerates. Finding the point where it breaks locates the corruption to within a byte, which is far more useful than knowing the whole string is wrong.

Question marks and replacement characters

A string of U+FFFD replacement characters means the bytes are valid but do not form legal sequences in the encoding being applied. Usually the data is UTF-8 being decoded as Latin-1 or the reverse. The bytes are intact; the interpretation is wrong. Trying a different encoding costs nothing and fixes it most of the time.

Separators and whitespace

Binary is pasted with spaces, without them, with newlines every so often, or in comma-separated groups. A decoder should strip everything that is not a 0 or a 1 before grouping. When a decoder fails on input that looks correct, a non-breaking space or a stray tab copied from a web page is a frequent culprit, because it is invisible.

Leading zeros are load-bearing

An eight-bit byte must keep its leading zeros. 1000001 and 01000001 look like the same number but are seven and eight bits, and a byte-aligned decoder reading the first will consume a bit from the next character onward. Binary copied from a source that trimmed leading zeros will not decode.

Not everything that decodes is text

Binary may represent an image, a compressed archive or machine code, none of which produce readable output through a text decoder. Output that is mostly control characters and unprintable bytes usually means the data was never text — which is itself a useful answer.