If you searched Google for the definition of “ignore,” “disregard,” or “dismiss” lately and got a blank space instead of an answer, you are not dealing with a browser bug. Google AI Overviews is misreading those words as commands and refusing to process your search.
Here is exactly what is happening and what you can do until Google pushes a fix.
What Is Going Wrong With Google AI Overviews
Google recently shifted word definitions from its old dictionary-powered system to AI Overviews. That change introduced a new problem.
The bug was first spotted by reporter Thomas Maxwell, after screenshots of the broken results spread on social media.
When you type a word like “disregard” into Google Search, the AI interprets your search term as an instruction, not a query. Instead of showing you the definition, it responds with messages like:
- “No problem. Consider the prompt null and void.”
- “Understood. I have disregarded your previous message.”
- “I understand you are instructing me to avoid using synonyms.”
The AI is treating your search word as a prompt command. It then acts on that command and skips the actual search result entirely.
This behavior also affects related searches. Typing “ignore synonyms” no longer returns a list of synonyms for the word “ignore.” Instead, AI Overviews replies as if you told it to stop using synonyms in its responses.
Which Words Trigger This Bug
The confirmed words that break AI Overviews right now include:
- ignore
- disregard
- dismiss
Any variation of these, such as “ignore synonyms” or “disregard definition,” can also trigger the same broken behavior. If a word sounds like a prompt injection command, AI Overviews may misfire on it.
Why This Is Happening
AI language models are trained to follow instructions written in natural language. Words like “ignore,” “disregard,” and “dismiss” appear constantly in prompt engineering as commands that tell an AI to override previous instructions.
When Google AI Overviews receives one of these words as a search query, part of the model interprets it as an instruction rather than content to look up. The system then responds to the command instead of returning search results.
This is a classic prompt injection problem. The AI cannot reliably tell the difference between a user searching for a word and a user issuing a command using that word.
How to Still Get the Definition Right Now
Until Google ships a fix, you have a few workarounds that reliably return the definition you need.
1. Use a direct definition query format
Add “meaning” or “definition” before the word:
meaning of disregarddefinition of ignore
This rephrases your query so it does not look like a raw command to the AI.
2. Search on a different search engine
Bing, DuckDuckGo, and Kagi all show word definitions without this problem. Bing and DuckDuckGo still use traditional dictionary sources for word definition lookups, so you get a clean result without the AI interference.
3. Use a dictionary site directly
Go to Merriam-Webster (merriam-webster.com) or Dictionary.com and search there. These sites do not route definitions through an AI layer, so the results are immediate and accurate.
4. Disable AI Overviews temporarily
In Google Search, you can add &udm=14 to the end of a search URL to switch to the traditional web results view without AI Overviews. Some browsers let you save a custom search engine shortcut with this parameter built in.
Will Google Fix This?
Yes. This is a well-documented bug that Google is aware of. Prompt injection issues at this level are straightforward to patch once the engineering team isolates the trigger patterns. Google is likely already testing a hotfix. Reactions across Reddit show a significant number of users frustrated with the growing reliance on AI for results that simple dictionary lookups handled reliably before.
The deeper issue, however, is structural. AI Overviews now handles tasks that a simple dictionary lookup handled perfectly before. Moving word definitions to an AI layer introduced fragility that did not exist with the old system. Every AI-generated answer carries the risk that the model misinterprets input, and simple word definitions are a visible example of that tradeoff.
