Modern website — what the words in your proposal actually mean

Website proposals look similar today: next to the price sits a list of words there's no way to judge. Serverless. Edge. JAMstack. API-first. PWA. Sometimes WebAssembly.

None is inherently good or bad. Each describes a way of doing something, not whether it helps you. A good proposal and an expensive one differ exactly here: a real need behind the word, or one written in because it looks good.

This is a dictionary of those words for the person signing the contract, not writing the code: what each means, when it genuinely makes sense for your company, when it's overkill, and what to ask the contractor to settle it in one sentence.

What "modern website" actually means today

Worth setting a baseline, because "modern" in a proposal usually refers to looks, and decides something else entirely.

A modern website responds quickly on first load, works on a phone the same way it works on a computer, lets you change content without a developer, and doesn't fall over when ten times more people show up than usual. Four properties a customer actually feels, each checkable without taking anyone's word for it.

A fast first response. In your browser's developer tools, on the "Network" tab: throttle the connection to a mobile profile and reload. What matters is the server's first response and when the main content appears — not when the last script finishes. Run the same check on a competitor's site; a number with nothing to compare says little. How to measure so it means something is covered in our article on website testing.

A phone, same as a computer. Not whether the site "scales" — whether you can actually do what you came to do. Walk your purchase path on a phone, one thumb, weak signal: find the service, open the form, send it. Where you reach for a computer is where a customer gives up.

Content without a developer. Open the panel and count how many of these you can change yourself: homepage text, a menu position, a new post, the header photo, a price in a table. Whatever isn't on that list is a line on your invoice for minor edits for years.

Resilience to a traffic spike. What happens at ten times the usual traffic, and how do we know? "It'll hold up" with no load test is a claim, not an answer.

Every word below is a way of achieving one of these four things, not a goal in itself — so the question that settles a proposal isn't "is this modern", it's: which property does it specifically improve, and by how much. One caveat: some of these technologies improve things you don't need improved. That doesn't mean the contractor is tricking you, only building the way they usually build — the question above is enough to sort that out.

Six words from proposals, in order

Each one in the same layout: what it means, when it makes sense for you, when it's overkill. Ordered by how often they show up in quotes.

Serverless — you pay for execution, not for a machine

What it means. Code runs on demand, at a provider, with no server of your own to maintain — no machine sitting on around the clock, a bill for actual invocations instead.

When it makes sense for you. When traffic is uneven: a shop with a season, event sign-ups, a campaign bringing in two days' worth of a quarter's normal traffic. It also makes sense with nobody to administer a server, because there's nothing to administer.

When it's overkill. With steady, predictable traffic, ordinary hosting is cheaper and simpler.

The catch to settle in the contract. In a "pay for execution" model, the bill has no ceiling until somebody sets one. A loop that calls itself, or a bot hammering a form over a weekend, generates invocations just like real customers — except nobody notices until the invoice arrives. So ask: are spending limits and an alert set up, and what happens once the limit is hit — does the site stop, or does the bill keep growing.

Edge — content closer to the user

What it means. Instead of one machine in one city, the site is served from the node closest to whoever opens it — the CDN idea, stretched from files to code.

When it makes sense for you. Customers spread geographically: several markets, another continent, an app used while travelling. The difference in first-response time is then noticeable with no measurement needed.

When it's overkill. Selling in one country with the server in that country — the distance edge shortens is already short, and it adds complexity: code spread across nodes is harder to diagnose when something breaks for only some users. More on infrastructure in our piece on hosting and CDN.

JAMstack and static generation — a page ready before anyone arrives

What it means. Pages are built in advance, as finished files, served without assembling anything at visit time. Very fast loading, fewer moving parts to break — no database queried on every view, so nothing to overload.

When it makes sense for you. Content that changes rarely and there's a lot of it: a catalogue, documentation, a knowledge base, a blog, an offer page. The cheapest way to get a site that's fast always, not just at night.

When it's overkill — and one warning. Content changing hourly or depending on who's looking (per-customer prices, stock levels, a login) needs more mechanisms bolted on, adding back the complexity it was meant to remove. The warning is about the word, not the technology: it carries more search interest than the plain description of what it does — a sign it belongs to the contractor's vocabulary, not yours. Worth asking directly: how many minutes after a content edit does the change show up on the site.

Astro — a tool built for exactly this profile

What it means. A framework that builds sites so that the browser receives plain HTML, and scripts only load where something genuinely has to react to a click. In practice: less JavaScript to download and run than with general-purpose tools.

When it makes sense for you. A content-driven site with no heavy server-side logic — not our opinion, the research: in the State of React 2025 survey, Astro leads Next.js by 31 percentage points in developer satisfaction (92% versus 61% positive ratings among users; two years earlier it was 94% versus 85%), and the recurring complaint about the latter is growing complexity.

When it's overkill. When you're building an application, not a content site: a panel, a basket, a configurator, anything with a login and state. Then a content-focused tool starts getting in the way, and the conversation moves to choosing between Next.js and React.

API-first — one piece of content, many channels

What it means. Content and data are available through a programming interface, so the same product description can feed a website, an app, a screen in a showroom and a catalogue sent to a partner — without rewriting it in four places.

When it makes sense for you. When you genuinely have more than one channel, or know you will. This is also the foundation of headless architecture, which we cover separately.

When it's overkill. When there's one channel and nothing suggests a second. You end up paying for flexibility nobody uses, and managing content gets harder, not easier.

PWA — a website you can install

What it means. An ordinary website with three extras: it installs onto a phone's home screen with no App Store or Google Play, works offline within a scope you decide in advance, and can send notifications. The address stays the same, the content stays the same. A service worker handles this — a small program the browser keeps between the site and the network, able to answer even when there's no network at all (web.dev documentation, MDN).

When it makes sense for you. When a customer has a reason to come back: a panel with orders and invoices, order tracking, a catalogue for regular buyers, a task list for a field crew. An icon on the home screen has real value then, and installing from a browser is a far lower bar than downloading from an app store. Something proposals for mobile apps never mention: a PWA also installs on a computer — as a separate window with its own icon in the taskbar, no app store, no installer, no IT department involved.

When it's overkill. When a site gets visited once. A brochure site someone opens, finds a phone number on and closes doesn't need installing — nobody misses it.

Two things to settle before you get a quote. First, "works offline" isn't a switch — somebody has to decide exactly what's available with no connection, and remember the device will show the state from the last visit that had signal. The rule we apply ourselves: offline gets what changes rarely, everything else asks for a connection. Second, on iPhones, notifications only work after the site is added to the home screen, and are available from iOS 16.4 onward (the WebKit team's announcement). If notifications are a critical feature and your customers carry iPhones, this one condition can decide the whole thing.

How to check how much of this you already have. Open your own site on Android in Chrome and check the browser menu: if "Install app" shows up with your icon, the basic conditions are met. On iPhone, the same check lives in Safari's share menu — and only in Safari, worth knowing before you promise this feature to customers.

A limit worth knowing. A PWA stops where an app reaching into hardware begins: NFC, Bluetooth, background work with the screen off, sensors. A full comparison with a native app — with the arithmetic and the decision criteria — belongs to the decision to build an app, not a website; we treat that as a separate topic in its own right.

WebAssembly — and why it's probably not about you

What it means. A way to run code written in languages from outside the web world — C, C++, Rust — in the browser, at performance close to a program installed on the computer.

When it makes sense. For things that are computationally heavy: a graphics or video editor in the browser, a simulation, a 3D configurator, processing large files without uploading them to a server.

When it's overkill — treat it as a red flag. If WebAssembly comes up in a quote for an ordinary company website, there are two possibilities. Either the project genuinely has something that needs it — a 3D configurator, an editor, in-browser file processing — and the contractor will point to it by a single named feature. Or someone is testing a new toy on your budget. The question settles it in one sentence: which specific feature needs this. A general answer — "modernity", "performance" — is the second kind of answer.

Summary: which word improves which property

Four properties from the start of this piece, six words from proposals, and one table you can bring to a meeting.

The row that usually settles the conversation is the last one. If a contractor can say why the "overkill" case doesn't apply in your situation, you're talking to someone who has actually costed the project. If they answer "it's standard today" — you're getting a proposal written once for every client.

What these words won't do for you

None of them will sell your product.

A PWA won't get anyone to install your site if there's no reason to come back to it. Serverless won't lift your conversion rate. Edge won't help when a contact form has fifteen fields and the copy needs reading twice to make sense. Loading speed is a necessary condition — it isn't a sufficient one.

The order we apply ourselves and recommend to clients is the reverse of the order in most proposals: first the reason someone should come back, then the technical layer that makes that easier. If a decision about a PWA gets made before anyone has answered why a customer should return, it's made too early — and that question belongs to website strategy, not to a technology proposal.

Two words that appear in proposals and are not your website's technology

Two more turn up regularly in quotes alongside the six above: voice-search optimisation and artificial intelligence. Both are worth knowing how to set aside, because both look serious on a quote and neither improves any of the four properties this article is about.

Voice search. In English-language (UK) searches the phrase family looks substantial — around 720 searches a month. What settles it is not the total but the tail: google voice search, search by voice, googlevoicesearch. Those are not people looking for a company like yours; they are people trying to switch on dictation on their own phone. The share of that family asking how to prepare a site for spoken queries is a rounding error.

Artificial intelligence. Here the demand is real, and it is worth being precise about what it is demand for. In English-language (UK) searches the largest measured phrases are ai website builder, ai website creator and ai site builder, at 6,600 a month each, followed by free ai website builder (1,000) and wix ai (720). That is shopping for a tool that builds a site, which belongs to the tools section rather than to a decision about your own stack. The next block, ai seo at 1,900, is shopping for a service. Neither is a question about whether AI should sit on your company website — which is the thing a proposal is actually offering you.

The conclusion is the same as for the other six, and it comes down to one question: which of the four properties does this improve, and how do you know. If there is no answer, the line is in the quote because it looks good there.

A cheat sheet: questions to send a contractor

The questions below are written to be copied straight into an email, unedited. Answers in writing matter more than the questions themselves — a verbal assurance won't make it into the contract.

If the answers to the first three questions are specific, and the fifth gets "we'll set limits and an alert", you're talking to someone who has costed this project. If you get back phrases like "that's standard today" and "everything is flexible", you're getting a proposal written once for every client.

Where these numbers come from

• Keyword volumes — English-language demand (UK Keyword Planner pull), 11 September 2026, for the terms this article is actually built around: modern website 170, progressive web app 1,300, website technology 140. We did not pull volumes for the individual jargon words below (serverless, edge, JAMstack, API-first, WebAssembly) for this edition — they're covered here as terms a contractor might use, not as search targets in their own right.
• Voice search and AI volumes — a separate English-language (UK) Keyword Planner pull, not Polish-market volumes, 14 September 2026: voice search 720 with a tail of google voice search and search by voice; ai website builder, ai website creator and ai site builder at 6,600 each, free ai website builder 1,000, wix ai 720, ai seo 1,900.
• Astro's 31-percentage-point lead in developer satisfaction — the State of React 2025 survey, 3,760 responses collected between November 2025 and January 2026. Satisfaction is the share of positive ratings among users who gave an opinion, in the meta-frameworks section: Astro 94%, 94% and 92%, Next.js 85%, 75% and 61% in the 2023, 2024 and 2025 editions. Read at the publisher on 5 October 2026.
• The PWA mechanism — service worker, installation, offline mode — web.dev and MDN documentation, checked at source on 11 September 2026.
• Push notifications on iOS from version 16.4, after adding to the home screen — the WebKit team's announcement, checked at source on 11 September 2026.
• Cost comparisons between these technologies — we don't know of a study measuring them on comparable projects, so no multiplier is given here.