A WIRED report has identified a recurring failure in Google's AI Overviews, the AI-generated summary boxes that appear above organic results on Google Search: the system keeps telling users the year is 2024. The headline finding, "Google AI Overviews Says It's Still 2024," captures a defect that has shown up repeatedly in user screenshots and search experiments since the feature's broader rollout.

For searchers, the consequences are practical. An AI Overview that misstates the year can misframe everything from election results and policy changes to product launches and sports outcomes — the categories where temporal accuracy matters most. The problem recurs across enough query types that the WIRED piece treats it as a pattern, not a one-off glitch.

Google launched AI Overviews publicly at Google I/O in May 2024, expanding on the earlier Search Generative Experience (SGE, Google's testbed AI search interface) that had run in Labs since 2023. The feature pulls from multiple web sources and links out to publishers beneath the generated text. Marketers and SEO (search engine optimization, the practice of improving how pages rank in organic search) practitioners now track AI Overviews as a separate ranking surface from the classic ten blue links because inclusion or omission can swing click-through rates independently of blue-link position.

The failure WIRED documents appears tied to how the language model layer handles present-tense phrasing. Queries that read as time-sensitive — "who won the election," "latest iPhone," "stock price today" — seem to produce overviews that confidently stamp their own text with the wrong year. Google positions AI Overviews as drawing from its live index, but the model's outputs anchor heavily on training-data distributions rather than live signals, producing 2024 labels even when the underlying citations are recent.

What does this change for publishers?

Sites whose traffic depends on recency-sensitive verticals — news, finance, sports, tech reviews — already track AI Overview citations separately from organic rankings. A system that insists it is still 2024 erodes the comparative advantage of freshly updated stories inside AI summaries, because the model is less likely to surface the latest framing. Sites that invest heavily in same-day content may see less AI Overview pickup than the underlying index alone would predict, while evergreen pages retain disproportionate weight.

The pattern also has governance implications. Google's product documentation already tells users that AI Overviews can be wrong, and the company has rolled out disclaimers, source icons and a "show more" toggle that expands citations. None of those interventions directly address the underlying date bug. Users who trust the visible year stamp inherit the model's calendar error.

Google's public posture, as covered across multiple press accounts, is that AI Overviews improve continuously and that isolated errors do not represent the feature's average behavior. The relevant question is what "average" looks like at scale: a defect that appears in even a small share of queries produces a substantial count of bad answers across the billions of searches Google handles each day.

What to monitor next

The next update to Google's AI Overview documentation and any shift in how Search Central (Google's channel for webmaster communications) discusses date freshness. Publishers running controlled experiments — identical content, two timestamps, tracked AI Overview inclusion — will surface the cleanest signal first.