What does a typical website / app / automation project look like?

Every engagement starts with understanding the problem — what you need live, who it's for, and what “done” looks like. From there it's a tight loop: scope, build, review, ship. You'll talk directly with the engineer building it, not a project manager relaying messages. Most work lands in weeks, not quarters — with clear milestones and no black-box agency process.

What services does Hashton provide?

Three things: websites that convert, custom apps and tools for how your business actually runs, and AI automation that takes manual admin off your plate. You can book a single landing page or a fuller build — the stack and process stay engineer-led either way.

Do you use no-code builders or custom code?

Custom code by default — typically modern stacks like Next.js — so the result holds up under real use, SEO, and change. No-code tools are fine for throwaway experiments; if you need something credible and maintainable, we build it properly.

Can you help launch something quickly (MVP / landing page)?

Yes. The focus is shipping the smallest version that actually works for customers or leads — not every feature on a wishlist. We cut scope hard, ship fast, then iterate once it's live.

Only if you need to edit content often without a developer. Many marketing sites are faster and cheaper as hardcoded pages until that need shows up. If you do need a CMS, we'll pick the lightest fit for your workflow — not a platform for its own sake.

Do you work fixed-price or flexible scope?

Fixed-price for clearly scoped websites and automations. Flexible / milestone-based when the product is still evolving. We'll agree the model on the call before any work starts — no surprise invoices.

Every site ships with solid technical SEO: semantic HTML, performance, metadata, accessibility, and indexing hygiene. Broader content/keyword strategy can be scoped separately if you need it — the foundation is always built in.

My site is slow / scores poorly on PageSpeed — can you help?

Yes. Slow sites are usually a pile of small issues — heavy images, too much JavaScript, blocking scripts, weak caching. We audit, fix the bottlenecks, and re-check Core Web Vitals so visitors (and search) feel the difference.

If it has no URL, search cannot find it

2 September 2026·Harry Ashton·engineering

A lot of small-business sites are PDFs wearing a domain. Pretty, slow, and invisible. Search engines and answer engines do not “look at” your hero animation. They fetch URLs. If the thing you are proud of has no address, it does not exist in that index.

A traditional UK shopfront with vintage wares in the window — A high-street frontage still needs a URL behind the glass. Photo: Valerija Timofejeva / Pexels

Modals are not pages

I have built case-study grids that opened in a dialog. They felt slick in a demo. Google got a single /proof URL and a pile of buttons. The individual stories — the ones you would actually want cited — 404’d. That is not a content problem. It is an architecture problem.

The fix is boring on purpose: one slug per piece of work, a real <a href> from the index, a unique title and description, and the important outbound links as crawlable anchors. The live pattern is on Hashton proof: EO Charging, Thrive Learning, CodePrepped, and the rest each have a path. If any of those still 404, that is a bug in the route, not an excuse to go back to modals.

What crawlers and chat-search actually use

Sitemap <loc> values that match the host users land on. Ours are https://www.hashton.agency/….
Canonical tags that do not fight your redirect. If apex 308s to www, every canonical should be www. Fighting that with apex metadata is how you get duplicate-host noise for months.
llms.txt as a plain-text map for AI systems. Hashton’s is public at /llms.txt — services, proof URLs, location, blog.
Schema that is true. Region is fine. An invented street or telephone is not. Missing NAP is better than fake NAP; citation sites will copy the lie forever.
Internal links a parser can see. Footer and proof pages should name related products with hrefs. A role="button" is a gesture, not a citation.

Black and white aerial view of Nottingham’s canal and rooftops — Nottingham from above. Photo: Altaf Shah / Pexels

Production means you can change it on a Tuesday

A brochure only the original freelancer can edit is a liability. I put marketing pages in the app and long articles in Postgres so a draft can sit unpublished, get a proper slug, and go live without a redesign. That is how this post ships: status draft until it should be public; no sitemap entry until published_at is set.

The same rule applies to services copy. If you add “AI automation” as a sentence on the homepage but never give it a URL, you are asking a crawler to infer a product you did not bother to name. Give it a page, a title, and a link from the thing people already rank for.

A short test before you pay anyone

Open your site in a private window. Disable JavaScript if you can. Can you still reach the service you sell, the work you are proud of, and a way to contact you? View source, search for rel="canonical", and check it matches the address bar after redirects. Fetch /sitemap.xml and click three URLs. If any of those fail, you do not have a website. You have a demo.

If you want that test run on your current URL, email me or book a slot. I will send back the list of URLs that 404, the host mismatches, and the pages that should exist and do not — without a pitch deck.

If it has no URL, search cannot find it

Modals are not pages

What crawlers and chat-search actually use

Production means you can change it on a Tuesday

A short test before you pay anyone

Tags

Need production software built?