Why AI Can't Actually Confirm You're Open Right Now

A standalone AI model does not have a clock. It does not know what time it is, and it has no live connection to your door. When someone asks "is this cafe open right now" and gets a confident answer, that confidence is not coming from the AI checking anything in real time. It is coming from how current and consistent the venue's own data already was, sitting there, before the question was ever asked.

This is the least intuitive part of GEO for hospitality operators, because it sounds like exactly the kind of thing AI should be good at. It is not, at least not directly, and understanding why changes where the actual effort needs to go.

Why the AI Itself Cannot Verify This

ChatGPT, Claude, Gemini, and most standalone AI models generate answers from patterns in text, not from live sensors or a connection to your point-of-sale system. Without a tool call to an external, real-time data source, they simply cannot know your current status. The "open now" answer an AI gives is, functionally, a lookup against whatever hours data it can retrieve, almost always Google Business Profile, layered with website schema as a secondary corroborating source.

This means Google Business Profile is not one input among many for real-time queries, it is close to the only input that matters. It is the primary source AI engines cross-reference for business hours, and it is the closest thing to a real-time source of truth that any AI system currently has access to. A venue that updates its GBP hours immediately when they change is not doing a nice-to-have maintenance task, it is doing the one thing that actually determines whether "open now" answers come back right.

AI Overviews, AI Mode, and the Local Pack Are Not the Same Thing

Part of the confusion here comes from AI Overviews and AI Mode getting talked about as if they are interchangeable, when they behave very differently for local, real-time-sensitive queries. AI Overviews appear for roughly 68% of local searches overall in one large study, but for a simple local-intent query like "tacos san francisco," they showed up only about 15% of the time, versus 92% for purely informational queries and 97% for hybrid-intent ones. Google's own local pack, the three-listing map module backed directly by live GBP data, remains the dominant answer for exactly this kind of query, by product design, because Google recognized that proximity- and time-sensitive local queries are better served by a system with direct access to live business data than by a language model summarizing text.

The practical read: for pure "open now" queries specifically, there is limited room to differentiate against what Google Maps already does, the live GBP signal wins regardless of what else a venue does. The actual GEO opportunity sits one layer over, in occasion, mood, and attribute queries, "date night," "good for groups," "halal," "clubbing tonight", where an AI is synthesizing descriptive content rather than doing a pure hours lookup, and where there is real room to be the recommended entity rather than an interchangeable option.

Seasonal Hours Need Their Own Handling

Ramadan hours, Bali's high-season extensions, holiday closures, these are exactly the situations where hours data goes stale fastest, because they are temporary by definition and easy to forget to revert. The schema property built for this, specialOpeningHoursSpecification, exists precisely so a venue can declare a date-bound exception without rewriting its entire standard schedule. Publishing a visible "current hours" note on the homepage itself, not buried on a contact page, gives crawlers a second, corroborating source that reduces the odds of a stale answer surviving in more than one place.

Two Edge Cases Worth Planning For

New venues face a specific version of this problem. Most AI models' training data has a cutoff date, which means a venue that opened in the last few months may simply not exist in an AI's knowledge yet, independent of how good its hours data is. The fix is aggressive initial indexing rather than passive optimization: claim and verify Google Business Profile immediately, submit to TripAdvisor and any active local listing platforms, distribute an opening announcement to local food media, and keep social presence active enough to generate indexable mentions. Within three to four months of opening, a venue with a genuinely complete digital footprint typically begins showing up in AI citations. Before that, it is largely invisible regardless of accuracy.

Ghost kitchens and delivery-only operations hit a different version of the same issue: no walk-in traffic to discover, but AI-driven discovery still happens through the identical path, someone asks for the best laksa delivery nearby and gets a named recommendation. The GBP address here may be a commercial kitchen rather than a guest-facing location, and that is fine, the listing still anchors the AI's knowledge of the brand's city-level presence, paired with LocalBusiness schema marked with a delivery service type and optimized GoFood/GrabFood listings, since delivery platforms are themselves indexed and cross-referenced by AI engines.

Why This Keeps Getting More Consequential, Not Less

Google's AI Overviews expanded Bahasa Indonesia support in October 2024, as part of a rollout to over 100 countries, and Google AI Mode began understanding and responding in Bahasa Indonesia specifically as of 12 September 2025. Combined with Indonesia's own AI adoption curve, 212 million internet users at 74.6% penetration entering 2025, and Indonesia posting roughly 85% ChatGPT growth, reportedly the fastest-growing ChatGPT market in Southeast Asia, the volume of queries where real-time accuracy actually matters is rising quickly, not staying flat. A gap between your GBP hours and reality was a minor inconvenience five years ago, when most guests still called ahead or checked a map app directly. It is a compounding discoverability problem now, as more of that checking happens inside an AI conversation instead.

There is a useful framing behind why hours and structured facts specifically carry this much weight: Google has described its own AI systems as increasingly "ranking entities, not websites." That shift means the question being resolved is not "which page ranks highest" but "which real-world business, with which verified attributes, is the right answer right now." Hours are one of the most time-sensitive attributes an entity has. Getting them wrong is not a content quality problem in the traditional SEO sense, it is an entity-accuracy problem, and it is judged on a much less forgiving timescale.

The Honest Limit Worth Stating Plainly

No amount of schema work will let an AI tell someone your current wait time, or guarantee it gets a last-minute closure right the moment it happens. The closest available signal for busyness is Google's own "popular times" feature, drawn from aggregated visit data, not from anything a venue publishes. Promising AI-surfaced live wait times to a client or in a proposal is promising something the current technology cannot deliver. The honest, and more useful, framing is that GEO minimizes the lag between reality and what is recorded somewhere machine-readable. It does not create real-time omniscience where none exists.

Frequently Asked Questions

If we update our website hours, does the AI know right away?

Not necessarily, and not reliably. Update Google Business Profile at the same time, since that is the source AI engines reference most heavily for hours, and treat your website schema as a mirror of GBP rather than the primary record.

Can we get AI to accurately state our current wait time?

No, not reliably with current technology. There is no live wait-time signal any AI model has direct access to. Google's "popular times" feature, based on aggregated visit patterns, is the closest available proxy, and it is not something a venue publishes or controls directly.

We just opened. Why aren't we showing up in AI recommendations yet?

Most AI models have a training-data cutoff, so a recently opened venue may simply not exist in their knowledge yet, regardless of how accurate your data is. Aggressive initial indexing across GBP, TripAdvisor, and local food media, plus active social presence, typically gets a new venue cited within three to four months.

Does this real-time problem apply to delivery-only ghost kitchens too?

Yes, with a slightly different fix. Anchor your Google Business Profile at your commercial kitchen address, use LocalBusiness schema with a delivery service type, and optimize your GoFood and GrabFood listings, since delivery platforms are themselves cross-referenced by AI engines the same way review platforms are.

Is there any way to guarantee 100% accurate "open now" answers?

No, and it is worth being upfront about that rather than overpromising. The realistic goal is minimizing the lag between a real change in hours and that change propagating across every AI-referenced source, not eliminating the lag entirely.

Real-time accuracy is the foundation this whole category sits on, alongside the structured menu and schema work covered in our piece on why the menu can't be a PDF anymore. For the fuller technical GEO framework, see Tessar Napitupulu's Cited or Silent: The Definitive GEO, AEO & AI Visibility Playbook, free to start at arfadia.com/resources/ebook-cited-or-silent, also on Apple Books and Amazon Kindle. For a GBP and schema-sync audit, see GEO for Restaurants, Cafes, Bars and Clubs.