Analyze texts the way a reader would: the whole thread, not the last ten messages
Clark is a text conversation analyzer that reads a thread in order — every line, who sent it, how long the silences ran — then writes a report where each finding points at messages you can check yourself.
Free previewNo card neededDeleted on schedule
Line by line, in order
Fictional sample
23:41, and the Last Message of the Night
Start with the night this group worked, because it did work, and it is the reason the rest of this report is not a complaint.
What is a text message analyzer?
A text message analyzer turns a conversation file into a written reading of it. Clark takes a plain-text thread, reads every message in order while keeping the speaker and the gaps between replies, and writes a report whose findings point at messages you can check.
- Accepted files
- A plain text file (.txt), or a WhatsApp chat export in .zip or .txt
- Format
- One message per line: a timestamp where the source has one, then the speaker name, a colon, and the message
- Length
- From a few hundred messages to a long archive; very long threads are read in sections
- Languages
- Any — quotes are kept in the language they were written in
What you get when you analyze text messages
A plain transcript looks like a wall of sentences. Read in order, with the speaker and the gaps kept intact, it carries structure. These are the things Clark reads for:
Who starts, who answers, who goes quiet
Reply order and response gaps are read as they happen. If one of you opens with a question nine times out of ten and the other answers, that asymmetry is a finding — counted first, then described.
The dry stretches, dated
A text conversation is mostly silence. Clark marks the long stretches — the week nothing was said, the month a thread sat unanswered — and reads what changed on either side, instead of averaging a year into one mood.
Tone shifts inside one thread
One-word answers, dropped emoji, punctuation that changes, jokes that stop landing: each is a measurable shift. The report says when the register changed and which lines show it.
Group threads read as groups
In a group text, two people are often talking to each other while the rest watch. Clark keeps the speaker attached to each line, so that exchange is not flattened into "the group", and a quiet member stays visible.
The lines each claim rests on
Every chapter carries quotes checked back against the file you upload. They appear as chat bubbles, in the language they were written in, with the speaker still attached — so a claim you dispute is one you can point at.
From a transcript to a written read
01
Get the thread into text
Get the conversation out of the app as text: a WhatsApp chat export, or the conversation copied into a document you save as .txt. Nothing is uploaded at this step.
02
Let Clark check the format
Drop the file into the third step and say where it came from, so the right parser runs. Clark checks that it can recognize the speaker pattern before anything is analysed.
03
Confirm who is who, then read
Confirm who is who — the names are cleaned up and mapped before reading — then let Clark work. Every message is read in order and the gaps are kept.
Every claim arrives with the lines behind it
A finding without a quote is a horoscope. Each one points at the messages behind it, rendered the way chat looks — who sent it, in a bubble, in the original wording.
yeah
Sam
you good?
Sam
that was a joke right
can we talk about last night
These quotes were written for this page as illustrations. They are not from any real conversation, and no line from a user's chat appears on this site.
What the report actually contains
Not "high-quality analysis" — a document you can read in one sitting and check line by line. These are the six parts every report is built from:
- An opening you can judge on its own
- Four to six paragraphs with no headings and no bullet points — the part that has to be worth reading on its own before you trust anything below it.
- Chapters divided by what changed
- Several chapters, each with its own claim, its own quote and its own reasoning — divided by what actually changed in the thread, not by equal slices of time.
- Quotes you can locate
- Each one was checked back against the file and is rendered as a chat bubble, in the language it was written, with the speaker attached. A quote that is not in your transcript is not supposed to survive that check.
- Numbers, and their limits stated
- Where the file supports it, counts are reported as counts: messages per person, the longest silence, the span of the thread. Where it does not — a participant who wrote six lines — it says the sample is thin rather than turning six lines into a verdict.
- Where the words and the data disagree
- When the words and the message data disagree — someone saying everything is fine in a thread they stopped replying to — the report names the disagreement rather than the politer reading.
- An ending that stays on the page
- The last chapter is addressed to you and names what the thread cannot answer. It is not therapy, legal, medical or financial advice.
Text message analyzer: common questions
What does an AI text analyzer actually do?+
Because the transcript keeps what a summary throws away: who spoke, in what order, and how long the gap was between two lines. An AI text analyzer that only counts words can tell you a thread was long; reading it in order is what surfaces the unanswered message and the thread that restarted after a year.
What format does the file need to be in?+
It needs a recognizable speaker pattern: the name, a colon, then the message, on the same line. Most app exports look like that already, and a WhatsApp export adds a timestamp in front. If no pattern is found, Clark stops.
Does it work for group texts?+
Yes. One-to-one and group threads both work, and a group is treated as a group: each participant is read separately, and an exchange between two members stays between those two. When someone contributed too little to support a conclusion, the report says so rather than inventing a personality for four messages.
Does Clark read my phone's messages directly?+
No. Clark reads the text file that arrives through the wizard. It does not open a phone's message database, does not use your number, and does not pull anything from a carrier. That text is deleted within 48 hours if you never order a report, and within 7 days of your most recent one if you do.
My thread is years long. Is that a problem?+
Long threads are read in sections and composed into one report with a single timeline, so a ten-thousand-message archive does not become a summary of a summary. What weakens a report is focus, not length: two hundred messages where you wrote six is thin evidence, and the report says so.
How much does it cost, and what is free?+
The opening and the first part of chapter one are free and need no card. The full report is a single payment of $9.90 for that one report — no subscription and no second bill when you re-read it.
Does the conversation have to be in English?+
English is not required. Timestamps in different date orders are parsed, messages in any script are read, and every quote keeps its original wording rather than being translated.
Put a text thread in front of Clark.
Upload the transcript, read the opening and the first half of chapter one for free, then decide. One payment of $9.90 unlocks the full report.
Analyze a text thread