Unspecified, really? cppreference's [C documentation][1] says that it returns zero. The [OpenGroup][2] documentation doesn't specify a return value when the conversion can't be performed. This recent [draft][3] of the ISO standard for C says that if the value cannot be represented (does that mean over/underflow, bad parse, both, neither?), then it's undefined behavior.
So three references give three different answers.
You could always use sscanf instead, which tells you how many values were scanned (e.g. zero or one).
[1]: https://en.cppreference.com/w/c/string/byte/atoi.html
[2]: https://pubs.opengroup.org/onlinepubs/9799919799/functions/a...
[3]: https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2310.pdf
The Linux man page (https://man7.org/linux/man-pages/man3/atoi.3.html#VERSIONS) says that POSIX.1 leaves it unspecified. As you found out, it's really something that should be avoided as much as possible, because pretty much everywhere disagrees how it should behave, especially if you value portability.
sscanf() is not a good replacement either! It's better to use strtol() instead. Either do what Lwan does (https://github.com/lpereira/lwan/blob/master/src/lib/lwan-co...), or look (https://cvsweb.openbsd.org/src/lib/libc/stdlib/strtonum.c?re...) at how OpenBSD implemented strtonum(3).
For instance, if you try to parse a number that's preceded by a lot of spaces, sscanf() will take a long time going through it. I've been hit by that when fuzzing Lwan.
Even cURL is avoiding sscanf(): https://daniel.haxx.se/blog/2025/04/07/writing-c-for-curl/
If your use case can have C++, then [std::from_chars][1] is ideal. Here's gcc's [implementation][2]; a lot of it seems to be handling different bases.
[1]: https://en.cppreference.com/w/cpp/utility/from_chars.html
[2]: https://github.com/gcc-mirror/gcc/blob/461fa63908b5bb1a44f12...