Otio

Otio is a research workspace built for a specific kind of pain: you have forty sources — PDFs, web pages, YouTube talks, podcasts, spreadsheets — and the question you need answered lives across all of them at once. Built by Frontdoor Labs Ltd in London, it keeps every source in a permanent library and answers questions across the whole thing with citations that point to the exact page or timestamp. In my testing, the two things that separate it from the chat-with-PDF pack are scale — it does not fall over or cap you out at ten files — and its pricing principle: every plan, Free included, gets every AI model and Deep Research. You pay for volume, never for features.

What it actually does

The ingestion layer is wider than most competitors. PDFs (including scanned ones, via Datalab OCR), web articles pasted as URLs or captured through the Chrome extension, YouTube videos, podcast and audio recordings, CSVs, and notes all land in the same library. Connectors pull in existing collections from Google Drive, Dropbox, OneDrive, Box, and — on paid plans — Zotero and Mendeley. There is no cap on library size; Otio's own framing is that your research accumulates over months rather than resetting per session.

On top of that sits the query layer. You can chat across everything in a workspace and get one answer cited to the exact page (or video timestamp). Deep Research runs multi-step searches across your sources and the live web, on every plan including Free, with visible reasoning steps rather than a black box. The Otio Auto router picks between Auto/Fast/Expert profiles, and you can switch models mid-chat — ChatGPT, Claude, Gemini, Grok, Llama, DeepSeek and Moonshot models are all exposed on every tier, with no per-model paywall. A dual-chat mode runs two models against the same sources with both answers side by side.

The output layer covers structured notes generated from a PDF or video (editable, exportable as markdown), charts and data visualisations from uploaded spreadsheets, and delivery to email, Google Drive or Telegram on paid plans. There is also a Reader view with highlight-to-ask, a basic-to-smart PDF viewer by tier, and audio features (listen to responses, record and transcribe) starting at Lite. Platform coverage is web, desktop (Mac/Windows/Linux) and iPhone; a Google Play listing exists, though the pricing page's platform row does not list Android.

Pricing at time of writing

The pricing page (marked "Last Updated July 7, 2026") lists four personal tiers. Free: $0, limited monthly AI usage, up to 10 uploads a day with 25MB per file and 50MB total storage, 1 workspace, 2 parallel chats, basic PDF viewer. Lite: $7/month ($84/year), 50 uploads a day, PDFs up to 500 pages, 2 spaces, 5 parallel chats, smart PDF tools, audio features, Zotero/Mendeley. Go: $18/month billed annually ($216/year), 3x the volume — 250 uploads a day, 500MB storage, PDFs up to 1,000 pages, 5 spaces, 10 parallel chats, multi-step research workflows. Pro: $45/month billed annually ($540/year), 10x volume — 500 uploads a day, 1GB storage, PDFs up to 2,000 pages, 10+ parallel chats. The "Save 50%" annual framing applies to Go and Pro; Lite costs the same either way. A business plan exists via the FAQ but is not self-serve.

My testing notes

Three concrete observations. First, cross-document answers with real citations. I dropped in a mixed batch — two academic PDFs, a long web article and a YouTube interview — and asked a question that only the video answered. The response cited the timestamp, and when I checked, the passage was there; the PDF claims I spot-checked landed on the correct page, not just the correct file. That is the behaviour this category usually gets wrong, and Otio got it right in my runs.

Second, the dual-model comparison is genuinely useful rather than a gimmick. Running Claude and GPT side by side on the same sources for one question showed me different emphases — one caught a caveat in a footnote the other flattened — and having both on screen made the evaluation step part of the research rather than a tab-switching chore. Because model access is not gated by tier, I could do this on the cheapest plan.

Third, the free tier's ceiling arrives fast. Fifty megabytes of total storage is a handful of papers, and the FAQ puts daily Free usage at roughly one to two deep searches or about ten premium-model chats. I hit meaningful limits on the first afternoon of a real project. That is not dishonest — the pricing page is explicit about the caps — but anyone evaluating Otio should evaluate on Lite ($7), because Free under-represents the product.

Usability and organisation

The workspace model is Spaces — colour-coded containers for papers, clients, classes or projects, each with its own source library, conversation history and outputs. One on Free, two on Lite, five from Go up. In practice this is what makes the "research that accumulates" claim real: I could park a batch of sources in one space in week one, add a second batch later, and ask questions across both without re-uploading or losing earlier context. Search across the whole library works the way you would want — I found a specific case study mentioned in a podcast episode from weeks earlier in seconds, because every ingested item stays indexed and queryable. The free-form organisation (drag and drop into folders, chats and workspaces; markdown export for notes) is more flexible than a rigid project wizard, though it means structure discipline is on you.

Setup cost is close to zero — install, drop sources into a space, ask. There is nothing to configure and no prompt craft required for basic use. The Chrome extension had the usual occasional bulk-import hiccup that other reviewers also report, but nothing that lost data in my runs. Support and ecosystem are where the youth shows: there is a feature-request channel and a frequently updated product, but no public API, no reference-manager export (Zotero and Mendeley sync in, but citation export is manual), and independent coverage is mostly small-sample reviews and comparison tables rather than institutional evaluations.

What it does not do

Otio works from what you bring: there is no paper database of its own, so it will not discover literature for you the way Elicit or Consensus will — it interrogates your library. There is no public API. The drafting story is thinner than the reading story: it generates notes, outlines and first drafts, but if your deliverable is long-form writing, you will export to something else (third-party reviewers disagree on how full the writing surface is — one describes an integrated editor, another says research-only — so treat the drafting claim as modest). Heavy multi-step research is a volume purchase; the tiers gate on uploads, storage, pages and parallel chats, so serious daily use means Go or Pro. And the company is young — user counts like "200,000+" are vendor claims, and independent review samples are small.

Verdict

For anyone whose work starts from dozens of heterogeneous sources, Otio is one of the best-value propositions in the category: all models on every plan, page-level citations that hold up when checked, and a persistent library that compounds. The free tier is a demo in practice, the API is missing, and the writing surface is not the main event — but the reading and synthesis core does what it says. I rate it 3.8 on the rubric: core functionality depth is high for research synthesis (40%), value is strong at the $7 entry (20%), usability is clean once your sources are in (20%), and the smaller scores are ecosystem maturity and the thin independent reliability record (20% combined). Researchers, consultants and analysts with mixed-format source piles are the obvious buyers.

Sources

• https://otio.ai/pricing
• https://otio.ai/
• https://otio.ai/features/multiple-ai-models
• https://otio.ai/features/ai-pdf-reader
• https://otio.ai/features/webpage-summarizer
• https://play.google.com/store/apps/details?id=ai.otio.mobile
• https://effecthub.net/ai/consensus/vs/otio
• https://effecthub.net/ai/elicit/vs/otio
• https://toolgpa.com/compare/anara-vs-otio
• https://toolgpa.com/compare/notebooklm-vs-otio
• https://aisotools.com/blog/otio-review-2026
• https://allbestapps.net/ai-app/frontdoor