How Much Does It Cost to Develop a Website in the UK?

In the UK, a brochure site typically costs around £3k-£8k; a content-heavy catalogue usually sits in the £8k-£25k band; and a custom web app or portal commonly lands at £25k-£60k+. Those are market ranges, not a Code23 quote. After 20 years and 350+ projects, we quote a fixed band after Blueprint, with fixed change scopes priced before work starts, and agentic delivery keeps that band at roughly half the cost of a traditional agency bench for comparable scope. What follows is what actually moves the number, what ChatGPT can and cannot replace, and how warranty and ownership work when you hire us.

The short answer: UK website cost ranges by project type

Think in project shapes, not in a mythical “average website”.

Project shape	Typical UK market range	What it usually means	How we price it
Marketing / brochure	~£3k-£8k	Clear IA, strong design, CMS editing, forms, analytics	Fixed band after Blueprint
Catalogue / content-heavy	~£8k-£25k	Many templates, search, structured content, migrations	Fixed band - content and IA dominate
Web app / portal features	~£25k-£60k+	Auth, roles, dashboards, integrations	Fixed band closer to product build than “a website”
Multi-site rollout	Programme pricing	Shared design system, many branded instances	Programme pricing - we have shipped 100+ branded sites in one rollout

We do not invent a comforting round number and call it “the average website”. We publish the commercial mechanics: a fixed quote after Blueprint, senior specialists directing AI coding agents so turnaround compresses - the same 5x delivery framing we use across the studio - without handing you unsupervised model output.

What each shape costs in calendar and risk, in our experience:

Marketing / brochure. One primary journey, a handful of templates, forms that hit CRM or email, analytics wired properly. The spend is mostly IA, design system discipline and CMS so your team can edit without calling us. Cheap-looking quotes often skip Harden - accessibility, performance budgets, security basics - and you pay later in launch week chaos.

Catalogue / content-heavy. Hundreds of pages or products, faceted search, related content, migration from a mess of legacy URLs. Content modelling and redirect maps eat more hours than the homepage mockup. Agents accelerate template production; humans still own the information architecture and the “what dies on cutover” decisions.

Web app / portal features. Logins, roles, dashboards, document uploads, third-party APIs. You are buying product engineering wearing a website label. Compare quotes only against other product builds - not against brochure agencies competing on page count.

Multi-site rollout. One design system, many branded instances on a multi-tenant rollout architecture. Programme pricing beats per-site copy-paste because the cost sits in the shared kit, not in reinventing navigation fifty times.

For local context see web design in Reading. For the build vs buy-agency decision, how to choose a web agency. Service hub: websites and apps.

What moves the price: the six factors that actually matter

Buyers ask for a number before they have a scope. Scope is six levers. Pull any of them and the band moves.

Scope of templates and templates’ complexity. Five unique layouts cost less than a component system that must flex across 40 page types. A homepage hero, a standard content page and a contact form is one job. A modular system with conditional blocks, personalised modules and campaign landing variants is another. We price the component surface, not the sitemap row count.
CMS depth. A thin marketing CMS is cheap. Editorial workflows, preview, localisation and role permissions are not. If marketing needs draft → review → schedule → publish with different roles, that workflow is product work. If three people need “edit anything” and nothing else, keep the CMS thin and spend the money on conversion design instead.
Integrations. CRM, booking, ERP, payment, SSO. Each trustworthy integration adds discovery and Harden time. A Zapier webhook is not the same as a bidirectional sync with error handling and idempotent retries. We map every integration in Blueprint: auth method, failure modes, who owns the third-party account, what happens when their API changes.
Content and migration. Moving years of messy pages is often the critical path. Agents help rewrite templates and bulk-transform structured fields; humans still decide redirects, canonicals and which orphan pages deserve a 301 versus a quiet death. Under-budgeted migration is the classic reason a “cheap” rebuild slips three months.
Design ambition. Custom illustration, motion and design-system work are real hours. Template skins are not the same product. If brand already has a locked system and you only need competent application, say so - that is a different band from inventing a visual language from scratch.
Assurance. Accessibility, performance budgets, security review, QA. Skipping Harden is how “cheap” launches get expensive. We treat Harden as in-scope delivery, not an upsell after the marketing site is already live and failing Lighthouse or WCAG checks in a sales demo.

App-shaped work sits in a sibling piece: app development cost UK. Redesigns have their own commercial pattern: website redesign cost.

Two quieter movers we see on almost every UK brief:

Stakeholder count. Five approvers with no decision owner inflate calendar time more than adding a template. We price the delivery system; politics still burns days unless someone on your side holds a single yes/no.

Third-party readiness. “We’ll get brand assets next week” and “the CRM sandbox is coming” are how fixed bands acquire change requests. Blueprint assumptions state what you supply and by when. Late assets do not magically stay free because agents are fast.

Template vs custom vs AI-assisted builds: what you get at each level

Template / theme. Fastest calendar time, lowest uniqueness, upgrade and security debt if you rely on plugin piles. Fine for proof-of-presence. Weak when the site is a sales engine. You inherit someone else’s information architecture and fight it every time your offer does not fit the demo. Themes win on day one demos and lose on year-two change velocity.

Custom. Architecture matched to your content model, performance and editorial needs. Ownership is clear. Historically this was the expensive lane - large benches, long discovery, weeks of hand-written UI before anyone saw a working CMS. Custom still wins when search, auth, multi-brand or unusual workflows are the product, not a plugin.

AI-assisted custom (how we work). Same accountability as custom - senior people lead - with agents compressing implementation. You get cleaner iteration speed and the half-cost band versus a traditional bench, not a slot machine of random page generators. Quality gates stay human: review, Harden, release authority.

How that lands week to week in our studio:

Seniors set architecture, content model and non-negotiables before agents touch production paths
Agents draft components, tests and refactors against that frame at the 5x delivery pace we use across builds
Humans cull options, enforce accessibility and performance budgets, and own go-live
Extras outside Blueprint stay on transparent fixed terms - speed does not turn scope creep into a free lunch

If someone sells “AI website” as unsupervised generation with no Harden phase, you are buying a demo. If they sell agents inside a senior delivery system, you are buying leverage.

A practical buyer test: ask who signs the release, what the warranty covers, and whether the codebase is yours. Est 2005, 350+ projects and 60+ five-star reviews did not come from shipping unreviewed model output into client hosting accounts. Agentic mastery means seniors wield agents as an unfair advantage - not as a substitute for a named team.

When comparing quotes, line up the same level. A theme install against an AI-assisted custom Blueprint is not a like-for-like tender. Compare theme to theme, custom to custom, and insist the AI-assisted lane still includes a fixed scope, a Harden pass and clear ownership terms.

Can ChatGPT build your website? What AI genuinely changes (and doesn’t)

Can ChatGPT actually create a website?

Yes - badly, for anything you would bet a business on. ChatGPT (and similar tools) can scaffold HTML, suggest copy and spit out component fragments. They do not own your hosting, redirects, CMS permissions, accessibility debt, payment flows, editorial workflow or the 3am incident. Agent-assisted delivery with human review is a different product: agents write and refactor under direction, seniors enforce architecture and ship. We show that workflow in the AI agent build log and the build-time dataset.

What AI changes in our builds:

Faster first drafts of components and tests
Quicker refactors across large codebases
More options explored before a human picks a direction
Shorter loops when a design token or API shape shifts mid-project
Less idle senior time waiting on boilerplate that agents can produce under review

What it does not change:

Who is accountable when checkout fails
Taste in IA and conversion design
Migration judgement
Client politics and scope control
Whether the CMS model matches how editors actually work on a Tuesday afternoon
Security boundaries, secrets handling and release discipline

The cost curve bends because idle senior time drops - not because judgement became free.

Where buyers get burned, in our experience: pasting a ChatGPT landing page into production hosting, wiring a form to an inbox, and calling it done. That artefact has no redirect map, no content governance, no Harden pass and no warranty. It also has no path into continuous change without rebuilding. The model was never the project manager, the accessibility reviewer or the person on the hook when Google Ads traffic hits a broken thank-you page.

Where agents earn their keep: greenfield implementation against a locked Blueprint, expanding a design system across many templates, generating tests, and chewing through repetitive refactors. That is the compressible slice behind the half-cost framing versus a traditional agency bench for the same scope. Politics-heavy programmes and ambiguous stakeholder goals compress less - the bottleneck is decision latency, not typing speed.

We still lock scope before implementation starts. Agents do not remove the need to decide what you are buying. They remove the need to pay a large junior bench to type it once the decision exists.

What a quote should include

We do not accept majority-on-completion. Blueprint work and senior capacity at kickoff are real line items, not a free pitch. Milestones should track working software against the fixed band - not slide decks - and the final payment should sit at Harden complete and launch readiness, not “we’ll invoice when we feel finished”.

A quote worth signing should state:

Fixed band and what is in / out of scope
Payment split and milestone definitions
Hourly rate for extras (transparent fixed change bands here)
Warranty (90 days on our builds)
IP / ownership
Hosting and post-launch support path (maintenance cost covers the retainer side)
Assumptions on content, assets and third-party APIs

If a quote is only a page count and a smile, it is not a quote.

After launch, Support & Growth sits on public tiers of £495 (Cyber Shield), £1,850 (Unlimited Growth) and £3,450 (Scale) a month when you want a named team on hosting, monitoring, patches and change capacity. The build invoice and the retainer are different products. Mixing them in one vague “care package” is how buyers lose track of what they paid for.

Ownership should be boring and explicit: you keep the code and content under the contract. We are not interested in hostage licences dressed as retainers. If a supplier will not state IP transfer in writing, treat that as part of the price - you are renting a dependency.

Change control stays adult. Scope added after Blueprint is estimated on transparent fixed terms or packaged as a change band before work starts. Agentic speed makes small changes cheaper in calendar time; it does not make undefined scope free. We would rather pause and re-Blueprint than silently burn the warranty margin on “quick adds” that become a second product.

How much does it cost to develop a website in the UK?

It depends on project shape - brochure, catalogue, web app or multi-site - and is quoted as a fixed band after Blueprint, with fixed change scopes priced before work starts. Agentic delivery keeps that band at roughly half a traditional agency bench for the same scope. There is no honest single national average that covers both a five-page brochure and a custom platform.

How much should I pay someone to create a website for me?

Pay for a fixed scope, clear ownership, a warranty and a maintenance path - not for the cheapest mockup. In our model that means Blueprint pricing, transparent fixed change bands outside scope, 90-day warranty, and optional Support & Growth from £495/month after launch.

How much should I pay to have a website built?

Budget against the job you actually need. A marketing site and a portal are different purchases. Ask for the six cost drivers above in writing, then compare bands like-for-like. Half-cost agentic delivery only means something when the scope side of the comparison is identical.

How much for a 10 page website?

Page count is a weak predictor. Ten marketing pages with one template are a different band from ten unique application-style views with auth and integrations. We price templates, components, CMS depth and integrations - then map that onto your sitemap. Bring the 10 pages to Blueprint and we will say which job you are buying.

When you want a fixed band instead of a guessing-game RFP, start at websites and apps. We will Blueprint the real scope first - the number comes from that, not from a brochure average.