Alexa, Where's the Traffic? Preparing Content for Screenless Search

Tens of millions of American households now talk to a cylinder on the kitchen counter. The adoption studies this year agree the curve is steep, and the pitches have duly arrived, as the February bot post-mortem predicted they would.

Before the vendors define this one, let us do what we do with every new surface. Start from the interaction pattern, because the pattern contains the entire strategy.

A voice query gets one answer. Not ten blue links. One response, selected by the assistant and read aloud. On the screenless devices, it is credited sometimes and sourced visibly never. There is no page two on a smart speaker. There is barely a page one.

Every content-strategy implication of voice search follows from that single fact, and most of the vendor noise around it evaporates on contact with it too. The question is not how to build a ‘voice presence,’ whatever the deck says that means. It is how content earns the position of being the one answer, for which questions, and whether that position pays. Those are answerable questions with unshowy answers, which is our favorite kind, so let us answer them.

The pipeline, demystified

Voice answers do not come from a separate voice index. They draw, demonstrably and heavily, from the machinery we have covered for years.

Google’s assistant reads from featured snippets, the extracted answer boxes our snippet coverage documented when they were a desktop curiosity. Those are themselves drawn from content structured to be extractable. Direct answers to natural-language questions, stated plainly in the opening sentences of a well-marked-up section, on pages whose authority the ordinary ranking systems already trust. The other assistants lean on their own knowledge sources and partnerships, but the structural logic converges.

Which means the preparation for voice is not new work. It is the intensification of existing work, with the history to prove the continuity. Hummingbird bet on conversational query understanding in 2013, and we said the bet would compound. We have argued since schema.org’s founding that machine-legible content wins the machine-mediated surfaces. And the question-shaped content advice, write to the questions people actually ask, in the phrasing they ask them, has been the content-strategy refrain since the long-tail pieces.

Voice is those three trendlines meeting a microphone. Write like people ask. Structure like machines read. Earn authority like always. The strategy survives the new surface intact because it was never surface-specific.

What real usage actually is, and what it makes first

Now the corrective the pitch decks omit. The usage research is consistent that real voice queries are dominated not by the exotic but by the local and the practical. ‘Is the clinic open.’ ‘Directions to admissions.’ ‘Call the box office.’ ‘What time does the pharmacy close.’

For the institutions we serve, that makes the unglamorous first investment obvious and cheap. Hours, locations, phone numbers, and the structured data that carries them. Correct and consistent across your site, your schema markup, and the business listings the assistants actually consult.

It is the least fashionable work in this month’s topic and the highest-probability voice interaction your organization will have this year. The prospective student in a moving car asking for your visitor center, answered correctly or not.

The audit fits in a week. The local-listing claims verified. The organization and location schema validated. The hours accurate everywhere they appear, including the December exceptions, which is when the queries spike and the errors embarrass. Do this before any conversation about the futuristic tier, because assistants that cannot learn your hours from you will not be quoting your thought leadership either.

The boundaries

• Branded skills and voice apps: premature for nearly everyone we serve. The invocation problem is the bot problem wearing a speaker grille. Users must know your skill exists, summon it by name, and tolerate its rails. The usage data on branded skills says they largely go unused after novelty week. The narrow exceptions, genuine recurring-utility cases, will be obvious by the February bot test. Script-able, high-volume, honestly bounded.

• Voice commerce: watch, don’t build. The purchase-by-voice numbers are a fraction of the projections being sold. And our readers’ transactions (applications, appointments, donations) want screens for good reasons. Review, confidence, and record.

• Being the read-aloud answer for questions you legitimately own: yes, now. The snippet-earning work above, aimed at the question inventory your search data and front-line staff already hold, per the FAQ-mining method from our content-strategy coverage. This is durable investment. It pays on screens today, on speakers as they grow, and on whatever machine-mediated surface arrives next. It is simply authority, structured.

• Measurement honesty. Voice attribution is largely invisible this year. The uncredited answer leaves no referrer. Treat the work as brand and service infrastructure measured by proxy: snippet ownership, local-listing analytics, direct traffic drift. Log the expectation now.

The pattern noted

This is the second machine-mediated surface this year, the February conversational post was the first, where the durable strategy turned out to be the same one. Structure your content for machine legibility. Answer real questions plainly. Earn authority honestly. Decline the branded-presence tier until the usage data justifies it.

That convergence is not coincidence. It is the shape of the next era becoming visible, in which growing shares of your audience’s contact with your institution are mediated by systems that read your content and speak for it, crediting you variably. The strategic asset in that era is being the source the machines trust and extract. Which is, conveniently, the same asset every previous era rewarded.

Search is learning to listen. Write like people ask, and we will keep tracking who the machines decide to quote. The Hummingbird piece closed four years ago by advising readers to bet on the conversational direction before the payoff was visible. The cylinders on the counters are the payoff arriving. The next surface’s version of this post is already predictable: same strategy, new microphone. Position for the pattern, not the product, and each arrival gets cheaper to meet.