1 Four kinds of work
A website that is doing real work has at least one of four things going on. A visitor can start something — a quote, an estimate, a service request, a registration. A visitor can check on an account — an order, a project, a balance, a status. A visitor can look something up — a record, a product, an item of inventory, an availability. Or a visitor can get advice or intelligence they could not have got on their own — a price for their situation, a sizing recommendation, an answer that belongs to the person asking.
None of those is a new idea. Banks have had check-on-an-account flows since the 1990s, retailers have had look-something-up since Amazon, and trades businesses have offered get-a-quote forms for almost as long as forms have been on the web. What is newer is that the cost of building a working capability has fallen far enough that a small or medium business can credibly run one — or two, or all four — without an enterprise budget. Industry-consensus.
Start something. The visitor lands with a question and a short deadline. A calculator, a booking flow, a structured quote form — each is a way to give a useful first answer in the same minute. A 2024 peer-reviewed paper out of Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi found that pages carrying real working content — numbers, sources, things a reader is paying attention to — were over forty percent more likely to be cited by generative answer engines than equivalent pages without. Verified at primary source. arxiv.org/abs/2311.09735 · Jun 28, 2024. The mechanics of why a good tool actually holds a person are the subject of a separate piece on this site; the short version is that a clear target, an instant response, and the felt sense of doing the work yourself land at the same moment.
Check on an account. A returning visitor finds what they left behind — the order, the project, the photos, the conversation. The site remembers who they are; the next thing they want to do is one tap away rather than a phone call. The component cost of standing up basic account state for a small site has fallen by roughly two orders of magnitude over the last decade, so the absence of a portal on a small business website is now more often a habit than a budget choice. Industry-consensus. A cluster piece on this is planned.
Look something up. The catalogue. A list with attributes — fields a customer can filter, sort, or search. Products with prices and stock; rooms with availability and rate; jobs with location and date. Conventional search and the generative answer surfaces both prefer structured records to free-flowing prose, because structured records can be quoted as records. Industry-consensus. A cluster piece on this is planned.
Get advice and intelligence. A small piece of working software that takes the visitor's situation and returns an answer fitted to it — what a project would cost given their inputs, what size of unit fits, what their property is worth in this month's market, what the next opening is. The answer is not generic. It belongs to the person who asked for it. Industry-consensus. A cluster piece on this is planned.
The four are independent. A custom-millwork shop might need a quote calculator and nothing else; a property manager might need account state and a searchable inventory; a clinic might need a booking flow and not much more. The umbrella claim is not that every business needs all four; the umbrella claim is that at least one of them, fitted to how the business already serves its customers, is the right scope for the build. Up at the development overview, the four are laid out as the choices a small or medium business owner can pick from.
Chapter 1, in one line. A working website does at least one of four things; the right one to build is the one that fits how the business already serves its customers.
2 Findability follows from doing real work
The argument so far has been about utility — the value of the four capabilities to a visitor who is already on the page. The case for showing up on Google in the first place is separate, and it leans in the same direction.
The single best-evidenced content lever for being found in conventional Google search is also the one with the most rigorous backing for the generative answer surfaces that now sit on top of it. The same peer-reviewed paper named above — GEO: Generative Engine Optimization, by Aggarwal and colleagues at Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi, presented at ACM SIGKDD KDD '24 — measured the effect of adding visible statistics, quotations, and source citations to the body text of a page. Their finding, verbatim, is that the addition of those elements "can significantly boost source visibility, with an increase of over 40% across various queries." Verified at primary source. arxiv.org/abs/2311.09735 · Jun 28, 2024. The authors are careful to note two things. First, the lift came from edits to visible body text, not from schema markup or metadata. Second, the figure measures citation share in synthesized answers, not click traffic. It is a finding about which pages get quoted, not about how many clicks they get.
Freshness is the other lever the evidence supports independently. A large-N study by Ahrefs — 16.97 million citations across seven generative answer platforms — found that pages cited by those platforms were on average 1,064 days old, against 1,432 days for pages ranking in Google's organic top ten. The cited pages were roughly a quarter fresher. Verified. ahrefs.com/blog/fresh-content · 2026. That direction is corroborated by other vendor datasets — pages updated within sixty to ninety days are cited more often — but those are single-source each and should be read as supporting evidence rather than independent confirmation. Industry-consensus on direction.
Conventional Google ranking and the answer-surface citation are also pulling apart. Across several large-sample independent studies, only about seventeen percent of AI Overview citations also rank in Google's organic top ten; a parallel figure from Ahrefs, on a different sample, is thirty-eight percent and falling. Industry-consensus. brightedge.com · Feb 12, 2026. The two surfaces are no longer working off the same shortlist. Building for one no longer guarantees the other; building for either rewards the same set of things — structured content, primary sources, current dates, real numbers a reader would want to see.
One forward-looking line, kept short on purpose: independent measurement by the Pew Research Center put AI-generated summaries on eighteen percent of US Google searches in March 2025, with vendor trackers measuring higher and the prevalence rising. Verified. pewresearch.org · Jul 22, 2025. Worth building for; not the headline.
A correction worth flagging in the same breath. Schema markup — the structured-data labels that SEO vendors talk a great deal about — makes a page eligible for rich-result formatting in Google search. It is not, per Google's own published documentation, a ranking signal in itself. The peer-reviewed finding above is explicit that the visibility lift came from body text, not markup. Verified. developers.google.com — Search Central, structured-data policies. Cheap insurance for eligibility, then; not the lever it is sometimes sold as.
Chapter 2, in one line. What lifts a page in conventional Google search — primary sources, visible numbers, current dates, structured records — is the same set of things that comes naturally to a website that is doing real work for the visitor in front of it.
3 Not every business needs every capability
The honest concession. The argument above is for the median small or medium business; it does not generalise to every one.
A meaningful share of small businesses run with no website at all. The most recent figure available, cited by Wix from Zippia data, puts the share of US small businesses without a website at roughly twenty-seven percent. Industry-consensus. wix.com/blog/small-business-website-statistics · 2026. The reasons given are revealing: in the same survey, twenty-seven percent of website-less owners said they did not think a website was relevant to their industry, and twenty-six percent named cost. A sole operator booked solid on referrals may need none of the four capabilities, and the thesis here is not that they should pretend otherwise.
The defensible claim is narrower. Most businesses that do need a website benefit from at least one of the four capabilities, not all of them. The capabilities are independent and should be fitted to how the business actually acquires and serves customers. A trades business benefits from a booking flow and a live-availability surface. A retailer benefits from a queryable product catalogue and an account that remembers an order. A consultancy benefits from a calculator that takes a prospective client most of the way to a number. A real-estate brokerage, a property manager, a fabricator — different customers, different questions, different one-of-four. The mistake is bundling.
The boundary on findability is the same shape. The peer-reviewed and large-N evidence for the freshness-and-structure case is strongest where customers are asking informational questions — services, considered purchases, business-to-business — because that is where generative summaries and informational query traffic concentrate. It is weakest for purely transactional or local-intent searches, where Google still routes the click through to a complete purchase or a phone call. Businesses on that end of the spectrum get more out of the interactive and live-data sides of the four than out of the findability case.
Chapter 3, in one line. The thesis is one of four, not four of four, and some businesses are best served by none — the question is which kind of work would change the most for the specific customer in front of this specific business.
4 Where this leaves a small or medium business owner
The four kinds of work — start something, check on an account, look something up, get advice or intelligence — define what a website is actually doing for the visitor. They are not a feature list to bolt onto an existing site. They are a way of asking, for one specific business, which of the four would change the most for the customer who lands on the page. One of them, well fitted, is enough to change what the website is for.
The studio has its own bias to declare here, and dropping the line would be dishonest: building working websites is the work this studio does. That is an interest the reader should hold against the argument while reading it. The supporting evidence above is sourced and dated regardless of who is writing it; the case for any specific capability for any specific business is a different conversation, and one worth having only after the business owner has already asked themselves the first question.
Chapter 4, in one line. Pick the one kind of work that would change the most for the customer; build that one well; the rest of the four can wait until the first one has earned its keep.