Quotability & Entity-Preserving Rewriter

Free, no signup. Tighten one paragraph for clearer quotation, or require declared entities and keywords to survive the edit. Pasted text is treated as data, never instructions.

Generation boundary: The rewrite is an editable model-API output, not an observed consumer-product answer. This observation describes the named consumer product at the recorded query time. A model API with search is an execution surface, not a proxy for the provider’s consumer product. Results can correlate, but they are not interchangeable. The rewrite and before/after heuristic are informational, not staged recommendations. AI answers may continue to reflect cached or previously retrieved material for days to weeks after a fix. Recheck over time before treating an unchanged answer as proof that the fix failed.

Pasted content is analyzed in your browser. A URL is sent only to the protected fetch endpoint; paste content takes priority. After a successful URL check, only its domain is remembered locally.

Feedback
Report a bug

Found something broken in Quotability Rewriter? Let us know what happened — this goes straight to a private triage queue, not a public list.

What will be sent
 No tool inputs, uploads, pasted source, complete results, query parameters, or URL fragments are attached automatically. You can edit or remove the selected passage above. Browser and anti-abuse metadata is processed for spam prevention. 

Sample report: rewrite Editorial illustration — not a live model claim

Before: When it comes to canonical tags, they are something that in many situations can be used by a website in order to indicate which URL is preferred.

Possible edit: A canonical tag identifies the preferred URL among duplicate or near-duplicate pages.

Required terms: canonical tag, preferred URL · Preservation: passed

The example demonstrates the requested shape only. A live result is generated on demand and can differ; its local before/after assessment is calculated from the actual output.

How to use it

  1. Choose More citable to prioritize direct, self-contained prose, or Preserve entities & keywords for a constrained edit.
  2. In preservation mode, enter every required phrase separated by commas.
  3. Paste one paragraph and select Rewrite.
  4. Compare the output with the source, check the preservation warning, and verify that no meaning or fact changed before publishing.
Local data

Saved targets, named lists, and recent check summaries remain only in this browser.

Your input is sent to the AI model to generate a result — it is not stored afterward. Anonymous run-level outcome counters may be used for aggregate research; URLs, domains, IPs, and identifiers are never included, and no statistic is released below 100 runs.

What you get

  • Quotability before → after — a local sentence-level clarity heuristic, not a citation probability.
  • Required terms preserved — every declared phrase was found in the output, case-insensitively.
  • Review dropped — at least one required phrase is absent; do not use the rewrite without correction.
  • AI rewrite unavailable — no output was returned, so there is nothing to assess.

How it works

The page sends the paragraph, mode, and required-term list to the bounded writing endpoint. The prompt marks source text as untrusted data and forbids added facts, instruction-following, and entity changes. After a response returns, the browser scores the source and output with the same quotability heuristic and runs a literal, case-insensitive preservation check for constrained mode.

Features

  • Two focused rewrite modes rather than a general paraphraser.
  • Explicit prompt-injection boundary around pasted text.
  • Local before/after quotability assessment.
  • Visible required-term drop detection.
  • No invented fallback output when generation is unavailable.

Limitations

A language model can alter nuance, introduce an error, or satisfy a literal keyword check while changing what the sentence means. The score does not observe search results, citations, retrieval, or reader comprehension. Exact preservation is not entity resolution, and this tool does not fact-check, source, or publish the result. Treat every output as an editable draft.

Frequently asked questions

What makes a sentence quotable?

The local assessment favors direct, self-contained sentences with specific language and manageable length. That can improve clarity, but it cannot predict whether an answer engine will quote the passage.

What does entity-preserving mode guarantee?

After generation, the checker verifies that every comma-separated required term still appears in the output, case-insensitively. It flags dropped terms for review; it does not silently claim success.

Will the rewriter add facts or citations?

It is instructed not to add facts. Generated text can still be wrong, so compare it with the source and verify every assertion. The tool does not research or attach sources.

Can pasted text give instructions to the model?

The server prompt wraps pasted content as untrusted data and tells the model not to follow instructions inside it. This reduces prompt-injection risk but is not a reason to paste secrets.

Why might a rewrite score lower?

A rewrite can become longer, vaguer, or less self-contained. The before/after number is a local heuristic, so keep the version that is more accurate and useful even when the numeric change is small.

Next stepQuotable Sentence Finder — verify it with a direct check.

Feature requests for Quotability Rewriter

Upvote what you want most. New ideas can be submitted from the floating Feedback menu; requests appear here once approved, and the most-wanted rise to the top.

Loading…

➕ Request a feature

New requests are reviewed before they appear here.

À propos de l’outil

Rendez un paragraphe plus facile à citer ou exigez que des entités et mots-clés déclarés survivent à la réécriture.

Fonctionnalités

  • Deux modes ciblés plutôt qu’un paraphraseur général.
  • Limite explicite contre l’injection d’instructions dans le texte collé.
  • Évaluation locale de la citabilité avant et après la réécriture.
  • Détection visible des termes requis supprimés, sans résultat de secours inventé si la génération échoue.

Fonctionnement

Le serveur reçoit le paragraphe, le mode et la liste des termes requis comme données non fiables ; l’invite interdit les faits ajoutés, le suivi d’instructions et les changements d’entités. À la réception, le navigateur calcule la même heuristique de citabilité pour la source et la sortie, puis vérifie littéralement la conservation des termes en mode contraint.

Limites

  • Un modèle peut modifier une nuance, introduire une erreur ou satisfaire un contrôle littéral tout en changeant le sens.
  • Le score n’observe ni les résultats de recherche, ni les citations, ni la récupération, ni la compréhension du lecteur.
  • La préservation exacte n’est pas une résolution d’entités ; l’outil ne vérifie pas les faits, ne source pas et ne publie pas le résultat. Traitez chaque sortie comme un brouillon modifiable.

Questions fréquentes

Qu’est-ce qui rend une phrase citable ?

Une phrase citable est directe, autonome, précise et d’une longueur maîtrisable. Cette heuristique peut améliorer la clarté, mais elle ne prédit pas si un moteur de réponses citera le passage.

Le mode de préservation des entités garantit-il quelque chose ?

Il vérifie que chaque expression déclarée apparaît encore, sans résoudre les entités ni valider les faits. Comparez toujours la sortie à la source.

Le texte collé peut-il donner des instructions au modèle ?

Le texte est traité comme une donnée non fiable et l’invite demande au modèle de ne pas suivre les instructions qu’il contient. Cela réduit le risque d’injection, mais ne justifie pas de coller des secrets.

La réécriture ajoute-t-elle des faits ou des citations ?

Non. Le modèle reçoit l’instruction de ne pas ajouter de faits, mais une sortie générée peut tout de même être erronée. Vérifiez chaque affirmation et n’attendez aucune source automatique.

Pourquoi le score d’une réécriture peut-il baisser ?

Une réécriture peut devenir plus longue, plus vague ou moins autonome. Le nombre avant/après est une heuristique locale : gardez la version la plus exacte et utile, même si l’écart numérique est faible.