Where did I fail in the Wordle cracker I wrote in Python?

I wrote a program that helps me crack Wordle puzzles. However, I guess it is biased on the alphabetical order because of the site I scraped the word list from. When "audio" has no matching letters, my program bugs and doesn't give me any words (even though it has "jewel" in its word list). How could I fix it?

I used this piece of code for getting all the words Wordle uses, as well as creating a list of used letters.

import requests
from bs4 import BeautifulSoup

webpage = 'https://www.wordunscrambler.net/word-list/wordle-word-list'
webpage_response = requests.get(webpage)
content = webpage_response.content
soup = BeautifulSoup(content, 'html.parser')

words = {}
for a in soup.find(class_='content').find_all('a'):
    text = a.get_text()
    if len(text) == 5:
        words[text] = [letter for letter in text]

with open('wordle_words.py', 'w') as file:
    file.write('words = ' + str(words))

I won't share the output, since the list is big. I also made a algorithm for determining which word was the best opener, but I won't share it because it's not important.

This piece of code helps the user play Wordle.

from wordle_words import words

starting_word = 'audio'

print('Start by entering the word "audio".')

wrong_letters = set()
matching_letters = set()

for i in range(5):
    
    attemped_word = input('What word did you try? ')
    feedback = input('What were the matching letters? Uppercase for green letters, lowercase for yellow letters and underline for gray ones. "A_Do_": ')

    wrong_letters.update(set([letter for letter in attemped_word.lower() if letter not in feedback.lower()]))
    matching_letters.update(set([letter.lower() for letter in feedback if letter != '_']))
    matching_words = {word: letters for word, letters in words.items()
                      if matching_letters.issubset(set(letters)) and wrong_letters.isdisjoint(set(letters))}

    best_fit = ''
    best_match = 0

    for word, letters in matching_words.items():
        matches = 0
        
        for index in range(5):
            if feedback[index].isupper():
                if letters[index] == feedback[index].lower():
                    matches += 1
                    
            if feedback[index].islower():
                if letters[index] != feedback[index]:
                    matches += 1
                    
        if matches > best_match:
            best_fit = word
            best_match = matches

    print(best_fit + '\n')

I expected it would output "jewel" after I told it "audio" hadn't returned me any matches.

There is a flaw in this part:

for index in range(5):
    if feedback[index].isupper():
        if letters[index] == feedback[index].lower():
            matches += 1
            
    if feedback[index].islower():
        if letters[index] != feedback[index]:
            matches += 1

The problem is that an underscore is neither lower or upper. You need an additional branch for missed letters, and that should match += 1 always for that position.

for index in range(5):
    if feedback[index].isupper():
        if letters[index] == feedback[index].lower():
            matches += 1
            
    if feedback[index].islower():
        if letters[index] != feedback[index]:
            matches += 1

    if feedback[index] == "_":
        matches += 1

There is actually another flaw that is not mentioned in your question. You are incrementing the matches value and getting the better one, but this can give you unmatched words anyway.

For example, if you guessed "audio" and got "A_Do_", it might suggest you "today". It would only be a problem if the actual word to guess is not in the word list (otherwise, that would have the better score), so maybe you'd never see that happen if you are scrapping right. Anyway, worth mentioning.