PREP vs. Ockem Proof: two bets on automation
Continual Engine’s PREP is the competitor most like us — both products bet that machines should do the mechanical majority of remediation. The differences are in what happens to the last mile, what you can prove afterward, and what you pay.
PREP is a cloud remediation platform claiming to automatically remediate “90% of the document with 95% accuracy,” with AI alt text, format conversions, and an outsourced services arm — priced by customized quote. Ockem Proof makes a different bet on the same automation thesis: no accuracy percentages, because averages aren’t a compliance posture. Every machine fix is independently re-verified, the judgment calls go to named humans in decision queues, a document cannot be certified while any check is unresolved — and the price is flat, per document, published.
Side by side
| Continual Engine PREP | Ockem Proof | |
|---|---|---|
| The automation bet | Claims to automatically remediate “90% of the document with 95% accuracy.” | No accuracy marketing — every machine fix is independently re-verified, and unresolved means uncertified. |
| Pricing | Customized quote — contact required. | One flat price per document, published. The free assessment prices your whole library. |
| The last mile | Built-in checker reports and your review in the tool — or their outsourced services team. | Decision queues route each judgment call to a named human; every decision is recorded in the attestation trail. |
| Deliverable | Remediated PDF plus accessibility reports. | Remediated PDF plus a certificate: per-criterion outcomes, named attestations, immutable version history. |
| Where the AI runs | Cloud platform. | Ockem-owned inference hardware — documents are never sent to a third-party AI API by default. |
| Beyond PDF | Format conversions to HTML, Word, text, and ePub — genuinely broader. | PDF-focused: audit, remediation, certification, and library monitoring for document corpora. |
Choose Continual Engine PREP if…
- You need format conversions — ePub, HTML, Word — alongside remediation, especially in publishing workflows.
- You need complex STEM and image descriptions at publisher scale, where their AI description work is a real strength.
- You want an established higher-education vendor relationship with an outsourcing arm attached.
Choose Ockem Proof if…
- You need to know the price today, not after a quote cycle.
- “95% accuracy” isn’t something you can hand to a lawyer — you need per-document proof with names on it.
- Your documents can’t be processed through third-party cloud AI.
- Your volume is bursty — a backlog now, a trickle later — and self-serve flat pricing fits better than a negotiated platform deal.
Fair questions
PREP automates 90% of the work. Doesn’t Proof do the same thing?
The automation percentage is where the similarity ends. An average across documents is a marketing statistic; compliance is decided per document, criterion by criterion. Proof’s architecture assumes the machine will sometimes be wrong: every fix is re-checked independently, anything requiring judgment goes to a named human, and the certificate gate is pessimistic — one unresolved check and the document is not certified. The question to ask any automated vendor is not “what percent do you automate?” but “what happens to the part you got wrong, and who signed the result?”
Which is cheaper?
We can only tell you our number — one flat price per document, on the pricing page — because theirs requires a quote. That asymmetry is the point: you can compute your total cost with Proof before talking to anyone, and the free assessment will price your actual library. If PREP quotes you lower for the same corpus, bring us the quote; the comparison that matters is total cost to a certified, defensible document.
Where do our documents actually go?
With most cloud remediation platforms, your documents are processed on shared cloud infrastructure, and AI features may involve third-party model APIs. Proof runs inference on hardware Ockem owns, with self-hosted open models; a document reaches a third-party AI API only when every local backend is unavailable, and that fallback is off entirely when no external key is configured. Tenant isolation is enforced in the database itself. If your security review has specific language about data processing, bring it to us.
Competitor details, including quoted automation and accuracy figures, are taken from continualengine.com public pages as of July 2026 and may change. If we’ve got something wrong, tell us and we’ll fix it.