Building websites for the AI search era

For two decades, search meant a list. A visitor typed a query, received ten links, and chose. Increasingly it means an answer: a paragraph assembled by a model from the sources it trusts, with a small number of citations underneath. Whether that answer mentions your business, and whether what it says about you is true, now depends on decisions made in the structure and content of your website.

The encouraging part is that the things answer engines reward are, for the most part, the things a good website already does. The discouraging part is how few business websites do them.

What answer engines are looking for

A model assembling an answer needs to understand three things about a page: what entity it describes, what specific claims it makes, and whether those claims can be trusted. A page that is vague about who it is for, buries its facts in marketing copy, and contradicts the business's other listings fails all three. A page that states plainly what the business is, where it operates, what it does, and what it costs, and does so consistently, is the kind of source an answer is built from.

Write pages that can be quoted

Traditional search rewarded pages for containing words. Answer engines reward pages for containing answers. The practical difference is that a section of your site should be able to stand alone as a response to a question a customer would actually ask. "How much does a landscaping project cost in Wellington County?" should have a paragraph on your site that answers it directly, with a starting point and the factors that move the price, before it goes on to anything else.

• Lead with the answer, then explain. The first sentence under a heading should be the sentence a model would lift.
• One question per heading, phrased the way a person asks it.
• Specific over general. Places, scopes, timeframes, and starting prices are quotable; adjectives are not.
• Keep the facts in the text. Information that exists only in an image, a video, or an animation is invisible to the systems assembling answers.

Structured data that tells the truth

Structured data is a way of stating, in a machine-readable form, what a page is about: this is a local business, here is its name and area, this is an article, this is a question and its answer. Search engines have used it for years; answer engines lean on it harder because it removes guesswork. Every page on a business site should carry accurate structured data for what it is, and the data should agree with the visible content and with the business's other profiles.

The word that matters is accurate. Structured data that claims reviews the business does not have, or an address it does not occupy, is not a trick that works; it is a reason to be distrusted by the systems that check.

The technical foundations still matter

None of this replaces the fundamentals. Answer engines crawl the same web search engines do, and a site that is slow, blocks crawlers, hides its content behind scripts, or has no sensible heading structure is hard to read for both. Semantic HTML, fast loads on mobile, clean metadata, a sitemap, and a sane information architecture are the price of entry. We treat them as non-negotiable on every build for exactly this reason.

Content that earns a citation

Models are trained and tuned to prefer sources that add something. A page that restates what a hundred other pages say, in slightly different words, is unlikely to be cited over any of them. A page that says something first-hand, specific to your business, your area, and your experience, is a source with no substitute. The same discipline that makes content good for a human reader makes it citable: original observation, specific detail, a clear point of view.

This is also the strongest argument against filling a site with generated text. Volume is not the signal. Distinctiveness and accuracy are, and generated filler is the opposite of both.

For local businesses: the site and the profile must agree

For a business that serves an area, the answer a customer receives is assembled from several sources at once: the website, the map listing, the review platforms, the directories. When those disagree about the name, the hours, the phone number, or the services, the disagreement is resolved by leaving the business out. Keeping every listing consistent with the site is unglamorous work and it is where local visibility is won or lost.

What not to do

• Do not publish generated articles at volume. They dilute the site's distinctiveness and are increasingly identifiable.
• Do not stuff pages with question-shaped headings that lead nowhere. The heading has to be followed by a real answer.
• Do not mark up claims you cannot support. Structured data is a statement of fact, and false statements have consequences.
• Do not hide the substance behind interaction. If a visitor has to click, scroll, or wait to find the fact, so does the crawler.