What changed since the voice search hype cycle
The 2017–2019 voice search hype was based on projections that voice would be 50% of all searches by 2020 and that businesses needed dedicated voice optimisation. Neither materialised in the form predicted. Voice queries did grow (Google Assistant, Siri, Alexa all process meaningful query volume) but the optimisation question merged into the broader conversational AI surface as those assistants got LLM upgrades.
In 2026 the relevant question isn't 'how do I rank in voice search' but 'how do I get cited when an LLM-powered assistant answers a question my customers might ask'. The answer involves the same work as AI citation optimisation generally: entity strength, answer-engine optimisation, structured data the LLM can ingest cleanly, and presence in the third-party sources LLMs cite (industry publications, structured directories, comparison sites).
What's actually still worth doing
FAQ schema and conversational long-tail content remain useful — not because they 'optimise for voice search' specifically, but because they perform across modern conversational AI surfaces, including AI Overviews, AI Mode, and ChatGPT/Perplexity citation. Treat them as part of answer-engine optimisation, not as a separate voice discipline.
What's not worth doing: agencies pitching 'voice search audits' as standalone services in 2026 are usually rebranding generic content optimisation. Decline the standalone offering and invest the same effort into answer-engine optimisation, which covers the conversational AI surface comprehensively.