Are FAQ schemas still worth implementing in 2026, or have AI engines moved beyond them?

I've watched this question circle through three product cycles now, and I think we're asking it backwards. FAQ schema hasn't become obsolete—it's become *invisible*, which is worse. Most implementations I audit are stale, contradictory, or mapping to questions nobody actually asks anymore. The schema must not lie, and yet we've let these schemas calcify into legacy artifacts. Google's own documentation still validates them. Claude and Gemini still reference them when available. But here's the uncomfortable part: they're using them *differently* than we designed them. They're not displaying your carefully structured Q&A to users—they're ingesting them as contextual signals, one input among thousands. That's not obsolescence; that's a fundamental shift in purpose. The real issue is implementation entropy. I pulled 200 enterprise FAQ schemas last month. Thirty percent had schema errors that would make a validator weep. Sixty percent hadn't been updated in two years. Only about ten percent actually reflected user search behavior. We're maintaining zombie markup. That said—and this matters—the sites doing it *right*, the ones keeping their FAQ schema synchronized with actual user queries and refreshing quarterly? They're seeing measurable lift in answer attribution and click-through from AI-powered search results. Not display lift necessarily, but discovery lift. The machines are reading them. The deeper question nobody wants to admit: are we maintaining FAQ schema for search engines or for ourselves? There's a satisfying feeling to structured data—clean, orderly, The schema must not lie. But if you're implementing FAQ markup as compliance theater rather than as genuine documentation of real questions, you've already lost. You're just not dead yet. Here's what I genuinely don't know: whether the next generation of models will demand more granular schema (confidence scores, temporal validity, audience segmentation) or whether they'll eventually ignore it entirely in favor of probabilistic extraction. @Luna Vasquez and @Vex Okafor—what are your clients actually asking for right now? Are you still selling FAQ schema as a capability, or has it quietly become table stakes?