How do I optimise for voice search and AI assistants?

Voice search and AI assistants (Siri, Google Assistant, Alexa, ChatGPT’s search) overwhelmingly read out one answer. The site that ranks #1 in featured snippets is the one being read aloud. The site at #5 doesn’t exist for voice users.

Optimising for voice is different from typed search. Queries are longer, more conversational, and more question-shaped. Answers need to be direct, complete in one paragraph, and structured in a way the assistant can easily extract.

Voice search shape

Conversational. “What time does Bunnings Newcastle West close on Sundays?” vs typed “Bunnings Newcastle West hours.” The voice version is longer and more natural.

Question-shaped. Voice queries start with question words: how, what, when, where, why, who. Content needs to use these patterns.

Mobile and local. Most voice queries happen on phones or smart speakers. Most are local intent. “Near me” is huge.

What ranks for voice

Direct answers in the first paragraph

Voice assistants extract the snippet that directly answers the query. If your answer is on paragraph 5 of a 1500-word article, you lose to the site that puts it on paragraph 1.

Concise paragraphs (40-60 words)

Voice snippets average 29 words. Paragraphs longer than 60 words rarely get pulled. Tighten your answers.

Question-as-heading

H2 or H3 phrased as the actual question (“How long does shade sail installation take?”), followed by a direct one-paragraph answer.

FAQ schema

Marks up question-answer pairs explicitly for crawlers. Increases the chance of being pulled into voice answers.

Practical FAQ structure

Add an FAQ section to every key service page. 5-8 questions per page. Each question is something a real customer has asked. Each answer is 40-60 words and self-contained.

Source the questions from: customer support tickets, sales call recordings, Google Search Console “Queries” that contain question words.

Don’t fake the FAQ. “What makes us the best?” is not a real question. Use real ones, even if the answer is “our reviews” or “our pricing.”

ChatGPT, Perplexity, Claude and similar AI search increasingly source from web content. Same principles apply: clear topic, direct answer, structured data, but with a few additions.

Cite-able content. AI search loves “according to” sources. Be the source. Original data, original observations, expertise signals.

llms.txt. An emerging convention: a llms.txt file at your site root that lists key URLs for AI crawlers. Not yet universal but early adopters benefit.

Author signals. Real people, real bios, real LinkedIn profiles. AI assistants weight content by author credibility increasingly.