Why Your Website Matters More in the AI Era, Not Less

Somewhere in 2024 the question "do I still need a website?" started sounding reasonable. Nobody clicks. The answer appears above the links. Social is where the people are. Quietly, though, the machines writing those answers were reading one thing more carefully than ever: your website.

This essay is an attempt to hold two apparently contradictory facts in one hand. Fact one: fewer humans visit websites than at any point in the modern web's history. Fact two: what is written on your website has more influence over your professional fortunes than it did when visits were plentiful. The resolution of the paradox is simple once you see it. The audience did not disappear. It changed species.

Do I still need a website in 2026?

Let us answer the search query directly, because people genuinely type it: yes, and if you are a consultant, adviser, founder, author or anyone else whose surname is part of their product, you need one more than you did five years ago. Not because visitors are coming back. Because the systems that now stand between you and every prospective client, employer and journalist treat your domain as the canonical record of who you are, and compose their answers partly from what they find there. A missing or threadbare website does not read as minimalism to a machine. It reads as an evidentiary gap, and gaps get filled by whoever else has published about you, or by your competitor's tidy, well-documented site.

The traffic fell. The visitors got better.

The numbers behind the panic are real. In the twelve months after Google launched AI Overviews in May 2024, the share of searches ending without a single click grew from 56 to 69 percent (SEO Sherpa's compilation of AI search statistics). Google now answers on the results page for roughly a quarter of searches via AI Overviews. The blue-link economy that justified a website as a traffic machine is genuinely contracting.

But look at what happens to the visitors who still arrive. Conductor's 2026 benchmarks, in the same compilation, put AI referrals at only about 1.08 percent of website traffic across ten industries, and then deliver the punchline: visitors referred by ChatGPT convert at 14.2 to 15.9 percent, Perplexity at around 10.5 percent, Claude at up to 16.8 percent, against roughly 1.76 percent for Google organic traffic. A person who lands on your site because an AI named you has already been briefed, compared and half-convinced. Your website is no longer the place where persuasion begins. It is where a nearly finished decision comes to confirm itself.

Fewer visits, wildly better visits. That trade only pays off, of course, if the machines doing the pre-selling actually name you, which brings us to who is really reading your pages.

Who actually reads your website now?

Watch a personal site's server logs for a week and the modern readership reveals itself: a procession of crawlers with purposeful names. OpenAI alone runs three, each with a different job: GPTBot collects pages for model training, OAI-SearchBot builds the index behind ChatGPT's search answers, and ChatGPT-User fetches pages live when a conversation needs them (OpenAI's bot documentation). Anthropic, Perplexity and Google run their own equivalents. Every one of them is a reader that arrives, takes notes, and later repeats what it learned to a human you will never see in your analytics.

This is the sense in which your website works when nobody is looking. The visit that matters is not the human one on Tuesday; it is the crawl that happened in March, which shaped the answer given to a buyer in July. We mapped where the engines look before naming anyone in Where AI Looks Before Recommending Someone, and your own domain is reliably on the itinerary. Whether to welcome or refuse those crawlers is its own strategic question, with a real trade-off between privacy and visibility; we walk through it in Should You Block AI Bots From Your Website?

Why isn't a LinkedIn profile enough?

The tempting shortcut deserves a fair hearing. A LinkedIn profile is free, discoverable and already where your peers are. But it fails the machine-reader on four counts. The format is enforced: every human on earth gets the same boxes, so nothing about the structure distinguishes you. The markup is theirs, not yours: you cannot attach structured data telling engines exactly who you are and which other profiles are yours. Access is at the platform's pleasure: walled gardens decide what crawlers see, and the terms change without consulting you. And the asset is rented: accounts get restricted, features get paywalled, reach gets algorithmically re-priced. None of this makes profiles worthless; they corroborate. But corroboration needs something to corroborate. The canonical record has to live somewhere you own the deeds, and machines behave accordingly: when engines resolve a person into an entity, the personal domain is the anchor the other profiles hang from.

What are the four jobs of a website now?

If the website is no longer a shopfront, what is it? Four jobs, each of which only your own domain can do properly.

Most personal websites fail at the third job. They are business cards: a headline, a headshot, three adjectives and a contact form. A machine reading that site learns you exist and nothing else, and machines do not cite existence. The engines lift pages that carry checkable substance, which is the entire argument of Publishing AI Actually Reads: depth on your own domain is what turns your name from a fact into an answer.

The AI-era website checklist

Translating the four jobs into an afternoon-by-afternoon build list:

1. An about page written in the third person, stating name, role, specialism and history in plain declarative sentences a machine can lift whole.
2. One canonical bio, reused everywhere, so every copy of you on the web casts the same vote.
3. Person structured data in JSON-LD, with sameAs pointing to your genuine profiles, so entity resolution requires no guesswork.
4. A body of work: articles or essays that answer the questions your buyers actually ask, phrased the way they ask them.
5. Proof pages: named testimonials, case write-ups, press mentions, each corroborating a specific claim rather than radiating vague excellence.
6. Freshness discipline: visible dates, updated when content changes, because stale pages lose the citation vote to current ones.
7. Crawler policy you chose on purpose: a robots.txt that reflects a decision about AI bots, not a default you never examined.
8. A contact route a distracted, pre-sold visitor cannot miss.

None of these items requires a designer, and that is rather the point. The machine is indifferent to your aesthetic. It is reading for identity, evidence and consistency, the raw material of the whole discipline of answer engine optimization. If you would like the gap between your current site and this checklist mapped precisely, that audit is part of our services.

A note on effort, because the checklist can read like a renovation. It is closer to a levelling survey. Most professionals discover they already own half of these assets, scattered across old pages and abandoned profiles; the work is consolidation, not creation. The about page rewrite is an afternoon. The structured data is an hour with a validator. The publishing habit is the only genuinely long game on the list, and it happens to be the one that pays humans and machines simultaneously. Sequence it that way and the website stops being a project and becomes a routine.

A worked example: two consultants, one with a business-card site

Here's a simple, hypothetical comparison. Consultant A has a beautiful, minimal website, a headline, a headshot, three adjectives, a contact form, nothing else. Consultant B has a plainer-looking site, but it carries a dozen substantive articles under their own byline, a clearly worded about page, and structured data linking their name to their genuine professional profiles. A prospective client asks an AI assistant to describe each consultant's approach before a first call. For Consultant A, the machine has almost nothing to work with beyond a job title, so it hedges or stays vague. For Consultant B, the machine can accurately summarize their actual point of view, citing specifics from their published work. The prospect walks into the call with Consultant B already primed and half-convinced, and walks into the call with Consultant A starting from zero. Same industry, similar experience level, wildly different machine-readable evidence, and that difference is entirely within either consultant's control to fix.

The quiet compounding

Here is the part the traffic-mourners miss. A website built for machine readers compounds in a way the old shopfront never did. Every article you add becomes retrievable evidence. Every consistent fact strengthens the entity. Every crawl refreshes the machine's copy of your record. And with ChatGPT at roughly 900 million weekly users by February 2026, more than double a year earlier (Similarweb's generative AI statistics), the readership consuming the machine's summary of your website grows without you lifting a finger. The website did not die when the clicks left. It stopped being a billboard and became a source. Sources, unlike billboards, get quoted.