Social media traps Google in ‘AI Overview’ spelling tests

Just over a week after Google rolled out a search overhaul that expands AI Mode and AI Overviews, social media users are using spelling challenges to highlight what they call persistent errors—while Google insists blue links remain, AI Mode isn’t the default,

For more than a decade. Google’s search results have been built on something straightforward: you type a question. you get links. you click. But in the last week. the conversation has shifted from clicking to typing again—because Google Search’s AI Mode now drops people into an interactive AI-powered search box where follow-up questions can be asked without leaving Google Search at all.

The change has landed with unusual force on social media. where users say Google is “breaking the internet” by leaning harder into AI answers that. in their view. still come with avoidable mistakes. To prove their point. they’ve turned Google’s AI into an internet-wide spelling bee—one that. in their posts. seems to stumble even on the most basic letters.

A viral test centered on a simple prompt: “how many letter Ls are in the word ‘google.’” The AI reply, according to the post, said there are two Ls—and then respelled the word to show where, misspelling “google” as “Goolle.”

Other users described similar failures. Users posted examples like “kangaroo” spelled with three Ps. and “magnificently” spelled with two Os—spelling problems that spread quickly because they look obvious on the page. One person said they kept rerunning the same experiment with different words. The results, they wrote, included “imaginary Es in ‘astronomical,’ Os in ‘heuristic,’ and a G in ‘pneumatic.’”.

The test wasn’t just about letters. Another user repeated the approach with words they chose on purpose and argued the AI reliably trips up when the words get more complex. “Just pick any word with 4 or more syllables,” they said, “and watch the AI falter. Seriously. You can just keep going and going.”.

A different commenter said the issue isn’t confined to syllable count and claimed random testing produces errors “about ⅔ of the time.” Their examples included “town” spelled with an H, “sing” spelled with an R, and “bottle” spelled with a K.

Spelling wasn’t the only target. Users also pointed to answers that they said swap “prostrate” and “prostate. ” repeat a single word in a loop. and devolve into endless strings of numbers. The concern running through the complaints is simple: if AI Mode is designed to reduce the need to visit websites beyond Google Search. then the accuracy of those answers matters more than ever.

Google says the direction of travel is still controlled—and that the mistakes users are spotlighting are being treated as a fixable problem. The company’s spokesperson told Fast Company that blue links will continue to show in search results and that AI Mode is not the default for Google Search.

The spokesperson also said AI Mode is easily accessible and that users can navigate back to the traditional results page. In that explanation, Google stressed that follow-ups behave differently depending on how a person chooses to proceed. “If someone chooses to ask a follow-up from an AI Overview. or selects the AI Mode button in the Search box. then that takes them to AI Mode. It doesn’t happen automatically—people have to choose to navigate to AI Mode,” the spokesperson said.

Google also said in a separate response to a TechCrunch article that first sparked the online frenzy that it would keep showing “blue links on the search results page in addition to AI responses.” That statement reinforced the same central point: AI answers are added. not supposed to replace the old path by default.

Even so, social media remained focused on how users are guided. One user wrote that the rollout seemed “softened so publishers don’t riot. ” but argued that the “default path still funnels you into the machine.” That critique fits the broader unease: people may not experience AI Mode as automatic. but they may feel it is hard to avoid in practice.

Google’s response to the spelling criticism drew a direct line between the errors and generative AI itself. “Generative AI can make mistakes with spelling. and this is an area where we’re working on making improvements. ” a Google spokesperson said. The company also argued that the larger picture is more reliable. “AI Overviews are accurate for the overwhelming majority of queries we see in Search. and we’re continually investing in updates that further improve the quality of the experience.”.

There’s a tension at the heart of this week’s dispute. Google is pitching a more interactive search experience—where questions can be handled inside Google Search itself. Social media is responding by asking whether that shift is safe when even simple spelling. they argue. can still go wrong. For now. the debate isn’t happening in a lab or a quarterly report—it’s happening one word at a time. in public.

Google Search AI Mode AI Overview blue links generative AI spelling errors social media backlash Fast Company TechCrunch search overhaul