Skip to content

Media Observatory · Tool

Analyze an article with Turing

Paste the link to an article, a post or a PDF. We tell you whether its language respects LGBTTTIQANB+ people and what could be written better.

2,364 articles analyzed from 832 media outlets and accounts · average rating 94 · used by 8 people

Which article do you want to analyze?

Paste the link and in a few seconds you’ll see its rating, what to change and how to write it better. It’s free and you don’t need an account.

Turing 2.1What's new? Reglas MyR IA

Works with Articles and blogs PDF X TikTok YouTube Facebook Instagram Threads

From social media, only public posts: Turing does not log in, so it cannot see private accounts, closed groups, stories or “friends only” posts. When it analyzes a post, Turing saves a certified copy with a timestamp and sends it to Internet Archive: if it is deleted later, there is proof that it existed.

Have the document? Upload a PDF

If the PDF comes from a web page (an outlet, an organization, an agency), also paste its link above: that way the analysis is added to that outlet’s statistics. A scanned PDF (images only) has no text to read.

We do not store your data: to count how many people use the Analyzer, an anonymous cookie (a random number) is enough. If the article is about LGBTTTIQANB+ issues, we save its link, a copy of the text and the analysis for media statistics. If it is about other issues, we only show you the analysis and save nothing. PDFs you upload are read on the spot and the file is discarded. If you sign in with your account, we save in your history (which only you can see and delete) the link, title and result of each analysis, never the text.

With your account: My analyses y batch analysis (up to 50 links, with a download for Excel). Were you sent an analysis with a reference number? Verify the certificate.

Cuando una expresión es dudosa (una cita, una crítica, un testimonio), enviamos esa oración, el titular y un poco del texto previo a una IA (Google Gemini, GPT-OSS de OpenAI y Claude de Anthropic, en los servidores de Google, Groq y Anthropic) para saber quién habla. Si la nota parece un caso de violencia y hay dudas, también le enviamos el texto de la nota para que ayude a nuestro equipo a revisarlo. Si la nota es del tema y tiene foto principal, la IA la mira para saber si expone el cuerpo o una agresión. Nunca enviamos tus datos ni los datos que Turing oculta, como el nombre registral o datos de menores.

Turing 2.1

How Turing works

Two layers work together: our rules set the rating and an AI reviews the doubtful cases. Turing also verifies what the article says and leaves a certified copy of what it read. It all happens in a few seconds.

Video · 6 minutes

Meet Turing

The heart of Memoria y Resistencia: it connects the Memorial with the Media and Legislative observatories. It analyzes, verifies, classifies, monitors, evaluates, reports and learns.

Watch it on YouTube
  1. 1 Paste the link

    From an article, a blog, a PDF or a post on X, TikTok, YouTube, Facebook or Instagram. If it can’t be read, paste the text.

  2. 2 We read the text

    We remove menus and ads. We separate what the outlet says from quotes, testimonies, footnotes and references.

3 Turing analyzesv2.1

Reglas de Memoria y Resistencia

169 reglas sobre nombre y género, términos que dañan, amarillismo, culpa a la víctima y discurso hostil. Se basan en el Protocolo de la Suprema Corte, CONAPRED, OMS y ONUSIDA.

It assigns the rating
AI second opinion

Logos are trademarks of their owners; they only show which AI we use.

Only on doubtful expressions: it checks whether it is a quote, a criticism or the outlet’s own voice. That way we don’t punish those who denounce.

It can only reduce weight
Fact-checks what it says

Turing compares the figures, laws and cases in the article, and what the organizations and legislators it quotes say, with the Memorial and the Legislative Observatory. It shows you the evidence.

Does not change the rating
  1. 4 Your result

    On a separate page, with just your article: the rating, what to change, how to write it better and what it says, fact-checked. If the AI reviewed something, we tell you what and why.

  2. 5 Certified copy

    We save what we read exactly as it was, with the date and time sealed by third parties. If someone deletes it later, the proof that it existed remains.

    How it works

Step 5 · Certified copy

What we analyze can't be erased

When Turing finishes, it backs up and certifies the post on its own, without you doing anything. So even if a politician or a media outlet deletes or changes it, it can be proven what it said and when. Deleting it does not improve its rating.

  1. Digital fingerprint

    We compute the SHA-256 fingerprint of what we read. If a single letter changes, the fingerprint no longer matches.

  2. Third-party timestamp

    An independent service (FreeTSA) signs the fingerprint with the date and time. No one, not even us, can change it afterwards.

  3. Public record

    The fingerprint is also anchored in the Bitcoin blockchain with OpenTimestamps, which anyone can check.

  4. Copy at Internet Archive

    We request a public copy on archive.org. If the site does not allow it to be saved, as happens with X, we make an alternate copy.

  5. Monitoring

    We check from time to time whether the post is still online. If it is deleted, the page says so and the copy remains available.

Each copy has a public certificate with a folio number, QR code and a printable PDF. It serves as proof that the post existed; it does not replace notarization.

How does the media talk?

The Media Observatory has every article that has been analyzed, each outlet’s and account’s classification, the most used expressions and the articles outlets corrected.

See the analyzed articles

How we rate

Every article starts with 100 points. We deduct for each harmful expression and add for what the article does well. We rely on the Protocolo OSIEGCS of the Supreme Court and on the journalist guides of LGBTTTIQANB+ organizations.

  • 85Very good
  • 70Good
  • 50Fair
  • 30Poor
  • 0Bad
Derogatory terms
Insults or words that demean people because of their orientation, identity or gender expression.
Name and gender
Calling the person by a name or gender that isn’t theirs, or treating their identity as a disguise.
Blames the victim
Explanations that justify violence or stigmatize the victim.
Sensationalism
Adjectives and morbid details that seek to shock rather than inform.
Hostile speech
Expressions that present LGBTTTIQANB+ people as a threat, an illness or a sin.
Imprecise terms
Outdated or incorrect words that have a precise alternative.
Gender and rights perspective
Naming the person’s identity, bias-motivated violence, the demand for justice and those who support it.
Inclusive language
Naming all people (“la ciudadanía”, “las y los vecinos”) instead of the generic masculine.

It is an automatic tool. It can get irony wrong, as well as quotes the article criticizes or very short texts. Use it as a guide and always review the full text. Expressions inside a verbatim quote weigh less, and those the article rejects (“it was not a crime of passion”) do not deduct.

2026 Memoria y Resistencia, an initiative of Nodo Cívico. The analysis is done on our servers; its criteria and code are not free to use.

Turing 2.1 · November 6, 2026

Turing 2.1: if a post is edited, its worst version counts

Turing keeps watch over the social media posts it analyzed. If the account edits the text later, it saves and certifies the new version, but the worst of its versions counts for the rating: correcting is acknowledged, but it does not erase what was said.

  • The article is marked “Corrected later” and explains what changed, when, and which certificate each version has.
  • Press articles stay as before: their most recent version counts.
  • Saved articles are automatically re-rated with v2.1.

Want to know more? Create your free account: we notify you of every Turing improvement and you keep your analyses.

Create my account See all updates

Analyzing the article

Getting ready…

0%

  1. Read the pageWe open the link and separate the article's text.
  2. Check the languageRespect, name and gender, sensationalism, victim blaming and perspective.
  3. Memorial: names and recordsWe check whether it names a person in the Memorial or whether it is a possible new case; if it is on topic, we add it to the statistics.
  4. Legislative observatoryLegislators, what they said and the bill it talks about. The AI decides on its own.
  5. StatementsWe look for the same statement in other outlets to confirm it.
  6. Certified copyWe save what we read with a digital fingerprint and a timestamp, and send it to Internet Archive: if it is deleted later, the proof remains.

AI second opinion

If there are doubtful expressions, one of these AIs checks who is speaking.

Video · 6 minutes

Meet Turing

What Turing is, what it does and how it brings together everything we do at Memoria y Resistencia.

You can watch it again whenever you want: it is next to the tool.