How to optimize for voice assistants and rank in spoken AI answers
Voice assistants now answer with the same kind of AI systems as their text apps. Google is replacing Google Assistant with Gemini on phones and home devices, Amazon's Alexa+ is built on large language models, and Siri can pass questions to ChatGPT when the user allows it. So optimizing for voice is mostly optimizing for those engines, plus the local business data that spoken answers lean on hardest: hours, phone numbers, addresses and services.
Two things make voice different. A spoken answer often names one source or none, so it builds recognition more than traffic. And the answer has to survive being read aloud, which rules out anything that only works as a table, an image or a "see below". The old voice playbook built around featured snippets and speakable markup matters much less than it did.
Listings and crawler checks to do first
- Update your three main listings. Google Business Profile, Apple Business Connect (which feeds Apple Maps and Siri) and Bing Places. Check hours, including holiday hours, phone number, service area, categories and the services or products you list. Spoken "is X open now" and "call a plumber near me" answers come from these, not from your website.
- Make your NAP match everywhere. Use one version of your business name, address and phone number on the site and every listing. The NAP consistency entry explains why mismatches cost you.
- Check crawler access for the assistants' parent companies. Applebot gathers data for Siri and Spotlight; Applebot-Extended only controls whether that data may train Apple's models, so blocking it does not remove you from Siri. RankWave AI's entries for Applebot, Applebot-Extended and Amazonbot list what each one does. Googlebot, and Google-Extended for Gemini, are covered on the Gemini page.
- Put hours, prices and phone numbers in text on your site, not only in images, footers built by scripts or embedded maps.
- Ask your own questions out loud. Try ten questions a customer would ask on a phone and a smart speaker. Write down what the assistant says, and whether it names you.
Write answers that survive being spoken
An assistant reading your content aloud cannot show a table, an arrow or a bolded warning. Write the key answer on each page so it works as speech:
- One or two short sentences that answer the question without depending on the next paragraph.
- Spell out what matters. "Open until 9 p.m. on weekdays and 6 p.m. on Saturdays" works aloud; a table of hours does not.
- No pointers. "As shown below" and "click here" mean nothing when heard.
- State conditions early. "For a gas boiler, a yearly service takes about an hour" beats a sentence that adds "gas boilers only" at the end.
- Say your brand name the way people say it. If it is often misheard, add the common spoken form in your About page copy so assistants can match it.
Read the paragraph aloud. If you stumble, a listener will too. This style also suits text answer engines, so nothing is wasted.
What changed from the snippet era
For years, voice optimization meant winning the featured snippet, because Google Assistant often read it aloud, and some publishers added speakable markup. Speakable has stayed a limited beta for news content in US English, and generative assistants now compose answers from several sources rather than reading one snippet word for word. Snippets and clear passages still help because they make your content easy to quote, but chasing position zero specifically for voice is no longer the goal. Structured data that describes your business, such as LocalBusiness schema with hours and contact details, remains useful because it matches the facts assistants need. Hydrogen SEO outputs Local Business schema site-wide from its Schema screen, so check that its hours agree with your listings.
Measuring voice and what stays out of reach
Voice is the hardest channel to measure. Spoken answers pass no referrer, and no assistant reports voice citations to site owners. Use indirect signals: calls, direction requests and website clicks in Business Profile insights; a rise in branded searches after you appear in answers; and your own monthly spoken test. The measurement page covers the text engines, which usually move first.
Much of voice is outside your control. Each assistant chooses its own data partners for local results and can change them. Many answers name no source at all. In-car and appliance assistants vary by manufacturer. What you control is the accuracy of your listings and the clarity of the facts on your own pages, which are the raw material every assistant draws on.
Common questions
Is speakable schema worth adding?
For most sites, no. Google has kept speakable as a limited beta for news content in US English. Accurate LocalBusiness schema and clear answers do more for a typical business.
Does blocking Applebot-Extended remove me from Siri?
No. Applebot-Extended only controls whether Apple may use your content to train its models. Applebot itself is what gathers data for Siri and Spotlight.
Why does my smart speaker give a competitor's phone number?
Assistants pull local answers from listing data. If your listing is unclaimed, out of date or missing the right category, a competitor whose listing matches the request better may be chosen. Claim and update your listings on Google, Apple and Bing.