Voice search optimization, minus the recycled advice

How voice results are actually chosen

A voice assistant has no room for ten results. It reads one, and it takes it from a small set of sources: a featured snippet or equivalent extracted passage, a knowledge graph entry, or a local business record. Winning any of those wins the voice answer as a side effect.

This is why almost every genuine voice-SEO tactic is a rebadged AEO or local-SEO tactic. That is not a criticism — it is the reason the work is worth doing, because it pays off in three channels at once.

What genuinely helps

• Question-shaped headings that match spoken phrasing — "how much does X cost", not "X pricing"
• A short, complete answer immediately under the heading, readable aloud in one breath
• Plain sentence structure: assistants read punctuation badly, and nested clauses become mush
• Accurate, identical business details everywhere — name, address, phone, hours
• Fast pages, because assistants abandon slow sources rather than waiting
• FAQ content that uses the words people say, not the words a brochure uses

speakable markup: ship it, expect nothing today

The speakable property marks the parts of a page suitable for text-to-speech. Google’s documentation limits its rich result to news publishers in specific locales, so a small business site should not expect a feature from it.

It is still worth including, for a different reason: the CSS selectors are an explicit statement of which passage on the page is the answer. That is a useful instruction to any extraction system, and it costs three lines of JSON-LD.

"speakable": {
  "@type": "SpeakableSpecification",
  "cssSelector": ["h1", ".speakable-answer"]
}

The myths worth dropping

"Voice is 50% of all searches"

A projection from 2016 that never materialised and gets re-quoted annually with no source. Voice is real and it is a minority of queries.

"Optimise for long-tail conversational keywords"

Half-true and usually misapplied. Spoken queries are longer, but stuffing a page with "hey google how do I" phrasing helps nothing. Write real questions and answer them.

"Voice needs its own content strategy"

It does not. It needs the same clear, factual, well-structured content, plus accurate local data if you serve a location.

If you serve a physical area

For local queries the assistant answers from business records rather than web pages, so accuracy in those records is the whole job: consistent name, address and phone across every listing, correct opening hours including exceptions, and a category that matches what people actually ask for.

For a service-area business without a shopfront, declare the service area rather than an address, and keep the same name and phone number everywhere. Inconsistency across records is the most common reason a local voice answer names a competitor instead.

Questions people ask about this

Does voice search need separate pages?

No. Separate voice pages duplicate content and compete with the originals. Improve the pages you have with clearer questions and shorter answers.

How do I know if I am winning voice answers?

There is no report for it. The workable proxy is tracking featured snippets and extracted answers for your target questions, since those are what assistants read from.

Is speakable markup worth adding?

Yes, at three lines of JSON-LD, as long as you expect no visible Google feature from it. Its value is telling any extractor which passage is the answer.

Who wrote this

Anas Bin Masud builds e-commerce sites and does technical SEO for businesses in the UK, Canada and Pakistan — fifteen live client sites, six of them stores taking real payments. The examples in these guides come from those builds and from the audit that rebuilt this site, not from a content brief. More about how I work, or read the case studies.