You type your old domain into a web archive and the pages are there: the homepage from one year, a blog post from another, a contact page whose stylesheet never got saved. Each capture is real. Put together blindly, they make a site that never existed.
That is the whole problem with restoring from an archive. The data is fine; the assembling is the job. Here is how Wayback Press does it, step by step, including where the AI is used and where a human signs off.
The pipeline at a glance
scan → snapshot choice → LLM review of the design and every page → junk and spam dropped → URL structure + 301 map → HTML or WordPress → human review → delivery
1. Scan
The scan reads which pages of the domain the archive holds and when they were captured. The free scan stops here, plus a preview: the old design, the URL structure, and a few sample pages, so you can see whether the template you remember is actually in there. It is a look, not the full download. The Internet Archive throttles heavy use, so a scan covers a limited number of pages.
If you would rather run the download yourself, the free downloader on GitHub pulls an archived site as HTML or WordPress. It has no AI in it. It works; it just takes the snapshots as they come.
2. Snapshot choice
Most pages have several captures. Some show the real page; some show a parking page, a “site suspended” notice, a half-loaded layout, or content from after the domain changed hands. The paid restore picks the best snapshot per page instead of one date for the whole site.
3. LLM review of the design and every page
This is the part free tools skip. An LLM reviews the design first (is the template intact, are the assets there) and then every page: is this the real content, is the layout broken, does it belong to the site’s topic. Layout breaks get fixed. Anything the agent is unsure about gets flagged rather than guessed.
My agent can be wrong, so his review is a filter, not a verdict. The verdict comes in step 7.
4. Junk and spam dropped
Archives keep everything, including what you would rather forget: thin tag pages, duplicate versions, injected spam pages from a past hack, content from a later owner. These are dropped, not restored. For a hacked live site, the cleanup side removes spam pages, rogue users and injected code, asks search engines to drop the spam URLs (removal requests plus a 410 status), and hardens the site.
5. URL structure and the 301 map
Old URLs carry whatever links and history the site had, so the restore keeps the URL structure. Where a page cannot keep its address, it gets a 301 redirect to the closest real page. The 301 map is part of what you receive.
6. HTML or WordPress
You choose the output. Plain HTML is simple and fast to host. If you want to edit and grow the site, we restore to editable WordPress: real posts and pages you can log into and change, not a static copy dressed up as WordPress.
7. Human review
A human reviews every job before delivery: the flagged pages, the dropped pages, the redirect map, and a walk through the result. Nothing ships on the agent’s word alone.
8. Delivery
You get the site (HTML files or a WordPress install/export), the URL list and 301 map, and a note of what was dropped and why. Any credentials you gave us for the job are used only for that job and removed after delivery. Turnaround depends on the job: 2–3 business days for most restores and repairs; larger or custom jobs are quoted.
What we need from you
- The right to the content. You confirm you own the site or may use the content you ask us to restore. This is in our terms, and a DMCA takedown process is published.
- Access, if the work is on a live site: hosting, CMS and Search Console as the job needs.
Common mistakes we try to save you from
- One date for the whole site. It drags in the worst capture of half the pages.
- Restoring everything. Spam and thin pages from the archive come back with the good ones.
- New URLs. Changing the structure without redirects throws away the old links.
- Expecting rankings back by default. We restore pages, redirects and index hygiene. Whether rankings return depends on Google. No guarantees, ever.
The same pipeline builds new sites
Not every project starts from an archive. The same agent pipeline that restores sites also builds them from zero, which is our niche website building service. This site is the proof: it was built end-to-end by that pipeline, with a human signing off at each gate. More on that on About Wayback Press.
How the content on this site is written and checked
Pages here go brief → research → agent draft → fact check → human review → publish. Search figures come from Ahrefs (US). Product, policy and legal facts are checked on the primary source and linked. What cannot be verified is left out or marked; no invented numbers, prices, quotes, clients or results. A human reviews every page before it is published.
The archive remembers everything. The craft is deciding what to forget.