.

Theoretically, it's true

But in practical...

Yes, something that runs in 26n is 26 times slower than something that runs in n.

However, when talking about big O notation, that doesn't make a difference. There exists a C>0 for which they are both always faster than C*n. This means that by definition, they are both part of O(n).

This is why big O isn't always the be all end all. It's applicability always depends or your application. If n+26 would run in a day, 26n would need 26 days, but a n^2 algorithm could very well need a year to accomplish the same and 2^n could take a millennium.
If you compare something up to infinity, "26 times slower" is insignificant and can be neglected.

Oh I see,
but still O(26n) is much worse than O(n+26)

No it isn't. indexOf is called a fixed amount of times. The length of the alphabet does not increase when the input length increases.

@XoRMiAS
You forget indexOf is O(n)

So total is O(n^2)

In terms of big O notation, this has the same performance as basically all other solutions: O(n).

BAD Performance