Data voids now get an AI answer anyway

A data void is a search query where demand exists but credible pages answering it do not, and Google now fills more of those gaps with an AI Overview instead of an empty or thin results page.

On October 12, 2026, Search Engine Roundtable reported that Google shows AI Overviews for queries that look like typos and for searches where it cannot match specific pages. In one example, the Overview told the searcher the input looked like a typo or random keyboard input and offered to help. In another, Google generated a response where a no-results page would once have appeared. Barry Schwartz noted the behavior has been around for a while without much coverage, and that AI systems rarely say they do not know.

The typo case is handled reasonably. The no-results case is the one worth thinking about, because it changes what happens on queries about things the web has not written about yet. For many brands, that includes their own newest product names.

#What Google said in 2024

When AI Overviews launched broadly, some of the most shared early answers came from this kind of query. In her May 2024 post, Google Search VP Liz Reid explained that "How many rocks should I eat?" was a question practically nobody asked before the screenshots spread, and that the Overview linked to satire republished on a geology software site. She described it as a data void and listed fixes, including better detection of nonsensical queries that should not show an AI Overview and limits on satire and user-generated content in responses.

The term itself is older. Michael Golebiewski and danah boyd defined data voids in a 2019 Data & Society report as queries that return little to no results, a gap that manipulators can fill with their own content. Their concern was search results. The same gap now feeds a generated answer.

Two years on, Google's posture has shifted from suppressing AI answers on thin queries toward producing one and phrasing it carefully. That can be the right call for a searcher who fat-fingered a query. It is a different situation for a query that is rare because it is new.

#Google already treats voids inconsistently

Google has long had a separate mechanism for these queries. A 2025 study by Ronald Robertson and colleagues examined Google's warning banners for low-quality, low-relevance and rapidly changing results across 1.4 million unique queries collected in three waves between October 2023 and September 2024. Banners appeared on about 1% of queries, the set of bannered queries changed substantially between waves, and low-quality banners stopped appearing in August 2024. Their models flagged 29 to 58 times more low-quality data voids than Google had bannered.

The useful point for practitioners is not the exact ratio. It is that Google's own handling of thin queries has been changing and inconsistent from one collection to the next, so how a given rare query is treated next month is not something to assume.

#Where brands meet a void

Most marketing teams think of data voids as a misinformation topic. In practice, a brand creates small voids constantly:

• A product name announced in a press release before its documentation page exists.
• A feature renamed in the app while the old name lives on in forums.
• Common misspellings of the brand or product name.
• A new pricing tier, error code, or integration that customers search for verbatim.
• Comparison queries pairing your product with a competitor nobody has written about yet.

When those queries had no good match, the worst outcome used to be a weak results page. A searcher could see that nothing official existed. Now the searcher may get a fluent paragraph assembled from whatever is nearby: an old version of the product, a competitor with a similar name, a reseller listing, or a forum guess. The answer looks just as confident as one built on your documentation.

Google's AI features documentation says supporting pages must be indexed and eligible to show with a snippet, with no additional requirements, and that AI Mode and AI Overviews may use query fan-out, issuing related searches across subtopics. Fan-out is what makes a void dangerous. A query with no exact match still has neighbors, and the system will search those neighbors.

#What to do about it

The practical answer is to stop leaving your own voids empty and to check what fills the ones you cannot close.

Publish before you announce. Ship the canonical page for a new product, tier or feature before the name appears anywhere public. Make sure it is indexed and snippet-eligible on launch day, since that is the stated bar for being a supporting source. A press release that outruns the documentation creates a void with your name on it.

Keep retired names alive. When you rename something, keep a page or a clear section that states the old name and what it became. Searchers and forums keep using old names for a long time, and a redirect alone does not put the old term in the text an engine reads.

List the queries only your brand can answer. Build a short set of verbatim queries: product names, misspellings, error messages, plan names, integration pairs. These are low-volume, so keyword tools often ignore them. They are also where an AI answer is most likely to be assembled from the wrong material.

Run them on a schedule and read the answer. For each query, record whether an AI Overview or AI Mode answer appears, which URLs it cites, and whether the facts match your own pages. A cited competitor or reseller on a query about your own product is a void signal. Re-run after launches and renames, because the void opens on the day the name goes public.

Check your preview controls. If a key page carries nosnippet, a restrictive max-snippet, or data-nosnippet on its main content, it may not be usable as a supporting source, and the void stays open even though the page exists. The same AI features documentation lists those controls and notes that changes take time to be recrawled.

Separate typo traffic from void traffic. An Overview that tells a searcher their query looks like a typo is harmless. One that answers a real question about your product from a third-party page is not. Tag them differently in any audit so the second kind does not get averaged away.

#The trade-off

Answering every query is a product decision Google is entitled to make, and for most typos it is a better experience than a blank page. The cost lands on whoever owns the subject of a thin query and has not written about it yet. In 2024 Google's stated fix for voids was to show fewer AI answers on them. The current behavior suggests the fix is now on the publisher's side: write the page before someone searches for it.