Help & Guide

Everything Dissekt can do, how to use each feature, and tips to get the most out of it.

What is DissektThe Clarity ScoreLanguage signalsSingle scanKeyword topicChrome extensionTelegram botAccess & limitsTipsFAQReferences

What is Dissekt

Dissekt is an information transparency tool. Its purpose is to help you see how information is constructed — the rhetorical techniques, framing, sourcing, and tone behind a piece of content — so you can read more critically.

Dissekt does not decide what is true or false. It surfaces how something is built and argued, cross-references existing fact-checks, and scores the credibility of sources. The judgment stays with you.

The Clarity Score

Every analysis produces a Clarity Score from 0.00 to 1.00 — a measure of how transparently the content is constructed. Higher is clearer and more straightforward; lower means more manipulation, weaker sourcing, or persuasive intent.

The bands: 0.65–1.00 High (transparent), 0.35–0.64 Moderate, 0.00–0.34 Low (heavily constructed).

The score combines three dimensions, multiplied together so a serious weakness in any one pulls the whole score down:

Construction — how it is built: rhetorical techniques, argument quality, completeness.
Verification — how supported it is: fact-checker consensus, source credibility, named sources.
Intent — what it wants: manipulation cues, tone, narrative framing.

Coverage. The score also reflects how much real evidence was available. When there are no fact-checks, no known source rating, or very little text, the result leans on text-pattern heuristics and is flagged as limited signal — treat those scores as rougher. Missing signals are excluded and the rest reweighted, rather than quietly pulling the score to neutral.

Tip: A high score is not an endorsement and a low score is not a debunk. A clearly written opinion piece and a heavily sourced report can both score well; the score is about transparency of construction, not agreement.

Single scan

The core feature. Paste text, a URL, or upload an image, and Dissekt analyzes that one item.

  1. Go to Analyze → Single scan.
  2. Paste text or a URL, or attach an image.
  3. Choose Brief (fast) or Detailed (deeper).
  4. Read the Clarity Score, detected techniques, cross-references, and source credibility.

For a URL, Dissekt fetches and extracts the article automatically. Some sites block automated access (paywalls) — if extraction fails, paste the text directly.

Tip: Use Brief for a quick gut-check, Detailed when a piece matters and you want the fuller reasoning and cross-referencing.

Keyword topic

Instead of one article, analyze how a whole topic is being covered right now. Dissekt fetches recent articles on your keywords, analyzes each, and aggregates a coverage report.

  1. Go to Analyze → Keyword topic.
  2. Type a topic (e.g. "5G health risks") and press Suggest.
  3. Dissekt proposes related keywords — tap to add the ones that sharpen your search.
  4. Pick Brief (more articles) or Detailed (deeper, fewer), then Analyze topic.
  5. Read the aggregate report: average clarity, dominant techniques, and a per-source breakdown sorted least-to-most clear.

Searches are limited to recent coverage (about the past month) and span general web sources — news, analysis, and blogs. Sources that block automated access are noted and excluded.

Tip: Add 2–3 related keywords rather than one broad term. "Modi, Adani, fraud allegations" returns sharper, more relevant coverage than just "Modi".

Chrome extension

Analyze anything as you browse. Right-click selected text, a link, or a page and choose "Analyze with Dissekt" to get an instant read in the extension popup. It activates only when you explicitly use it — no background tracking.

Telegram bot

Dissekt also runs as a Telegram bot, so you can send content for analysis directly from chat. Handy for checking forwarded messages and links on the go.

Access & limits

Daily scan limits reset at 00:00 GMT.

Free (no account): 3 brief + 1 detailed scan per day.
Free member account: 25 brief + 10 detailed per day, plus your Dashboard and API access.

Brief and detailed scans are counted separately. Keyword topic analyses count as one scan of the chosen depth.

Tip: An account is free and unlocks your Dashboard, saved scan history, and API access.

Tips for good results

• Give Dissekt enough to work with — a full article or a substantial passage reads better than a one-line snippet.
• For URLs that fail to load, paste the text directly.
• Read the dimensions, not just the headline score — a low score driven by Intent (manipulation) tells a different story than one driven by Verification (weak sourcing).
• Use Keyword topic to see patterns across sources, not just single items.
• Treat the score as a starting point for your own judgment, never the final word.

FAQ

Representation & language signals

Alongside the score, Dissekt tags language patterns it notices. These are descriptive observations, not verdicts — each tag shows the exact words that triggered it, and you decide whether it matters. They do not affect the Clarity Score.

Tip: A report shows only the tags that actually fired, so a clean piece may show none — that's normal, not a failure. These patterns are detected in English only, so non-English articles may show fewer signals.

Framing & credibility language

Doubt-casting language — words that quietly question credibility (“so-called”, “alleged”, “purported”).
Asserted certainty — verbs presenting contested points as settled (“revealed”, “exposed”, “confirmed”).
Heavily hedged — frequent softeners that distance claims (“may”, “possibly”, “arguably”).
One-sided framing — subjective intensifiers that assume agreement (“clearly”, “obviously”, “of course”).
Sweeping generalization — broad claims about an entire group.
Dehumanizing metaphor — threat/disaster language applied to people (“flood”, “swarm”, “infestation”).

Toxicity sub-types (shown when detected): Severe toxicity, Obscene language, Threat, Insult, Identity attack, Sexually explicit.

Tap any tag in a report to see the specific words that triggered it.

References

Dissekt's scoring draws on peer-reviewed research. Key sources:

📄Da San Martino et al., EMNLP-IJCNLP 2019Propaganda technique taxonomy📄Baly et al., EMNLP 2018News-source factuality & bias (MBFC basis)📄Wachsmuth et al., EACL 2017Computational argumentation quality📄Pavlopoulos et al., ACL 2020Context-aware toxicity📄Hutto & Gilbert, ICWSM 2014VADER sentiment analysis📄Recasens et al., ACL 2013Framing & epistemological bias📄Borkan et al., WWW 2019Toxicity sub-labels (Jigsaw/Detoxify)📄UNDP, Human Development Report 2010Geometric mean for composite indices

Still stuck? Contact us.