How it works
From the texts on your phone to a report you can file.
Three steps. The first happens on your phone — we handle the rest. Searching is free and instant; you pay only when you download the report.
1. Get your texts off your phone
On Android, the free SMS Backup & Restore app does it in a few taps. On iPhone, use the Messages app on a Mac or a low-cost export tool. Co-parenting apps like OurFamilyWizard have their own export buttons. Our free export guide walks through each option step by step. Tip: skip "Include media" when you export — photos and videos make the file ~100× bigger, and everything here works from the message text. Don't worry about cleaning the file up — we handle attachments, group threads, and messy formats.
2. Upload it — or paste a share link
Upload the file directly — any size, big files show a progress bar — or paste a Google Drive, Dropbox, or OneDrive link. We read it, organize every message, and delete your original file. A few minutes later, your messages are ready to search.
3. Search, pick, download
Search in plain language ("threats about taking the kids," "messages after I asked him to stop") or tap a ready-made category like threats or money. Every result shows the conversation around it. Pick the messages that matter and download your report — every message with its date, time, sender, and source, formatted for court. $99, one time.
Frequently asked questions
How do I get my text messages off my phone?
On Android, the free SMS Backup & Restore app exports your full message history in a few taps. On iPhone, you can use the Messages app on a Mac, or a low-cost export tool. Co-parenting apps like OurFamilyWizard have their own export buttons. Our free export guide walks through each option step by step.
What file types do you accept?
Android SMS Backup & Restore XML, iMazing CSV, and OurFamilyWizard PDF. If you have a different format, email evan@texttimeline.com with a sample and we'll add support — most exports we've seen are supportable.
How long does it take?
Most exports of a few thousand messages are ready to search in 1–3 minutes. A multi-year history of 30,000+ messages takes 5–10 minutes.
What happens to the file I upload?
It's read, organized, and deleted — we never keep your original export. What we keep is a private, encrypted index of your messages so you can search them and build your report. Message text stays only because the report has to quote and cite it; the file you uploaded is gone after processing.
How does search find messages that don't use my exact words?
Two kinds of search run at once: one matches exact words and phrases, the other matches meaning. So a search for "threats about taking the kids" finds messages like "you'll never see them again" even though no word matches. Every result shows the messages before and after it, so you read the conversation, not just the hit.
Does AI write the report?
No. The report is built from messages you selected, each quoted with its date, time, and sender. A short summary is generated only from those selected messages, and every sentence in it points to the specific messages it came from. Nothing is invented, and nothing outside your selection is used.
For attorneys and the technically curious
What's actually under the hood
TextTimeline is a retrieval-first system on a deliberately small stack: FastAPI on Cloud Run with FAISS and BM25 running in-process (no external vector database), embeddings from Vertex AI text-embedding-005, and reciprocal rank fusion combining the semantic and lexical result lists. Each message is enriched at index time with sentiment, threat-language, financial, custody, alienation, coercive-control, and substance-abuse flags — that's what powers the category quick-picks. An LLM (Gemini) is used only for query expansion and for the grounded summary over messages already selected for the report — never to read or classify the full corpus. Every claim in the deliverable cites message IDs, senders, and timestamps, and the report carries a source-provenance page (source filename, processing timestamp, message counts and date range) to support authentication under FRE 901; a SHA-256 checksum of the original file is recorded at upload. Derived artifacts are encrypted at rest with TTL auto-delete.
See it on a sample case
The demo runs on a realistic, fully fictitious custody dispute — the same pipeline that runs on your own messages.