Skip to content

How AI summaries work

Effective date: 8 October 2026 (2026-10-08) · Last updated: 8 October 2026

Every story in SwipelyNews says ✨ AI summary. Here's what that means.

The news comes from real outlets

Every story is based on an article reported and published by a news outlet. The outlet's name and logo appear on every story, along with a "Read on [outlet]" button that opens the full original article. AI doesn't report the news, and it doesn't invent stories.

What AI does

For each article, we use an AI model (Google Gemini) to:

  1. Write a headline of up to 90 characters, and a summary of 2 to 3 sentences (up to 320 characters) with the key facts: who, what, where, when and why.
    • The AI is told to state only facts from the article, in a neutral tone, in its own words.
    • It's told to keep claims attributed (for example "police said"), and not to add opinions.
  2. Pick photos or videos from the article that illustrate the story.
  3. Translate the headline and summary into the app's languages: English, French, Portuguese, German, Spanish, Japanese, Korean, Chinese and Arabic.
  4. Sort the story into topics, tags and countries, so your feed and Discover work.
  5. Flag graphic media (violence, gore, nudity, sexual content, self-harm or disturbing images). Flagged media is blurred until you tap to view it.

We also group stories about the same event from different outlets under Full coverage, so you can compare how they report it.

What AI can get wrong

AI summaries are generated automatically, and a person doesn't check each one before it appears. A summary can:

  • leave out important context or nuance;
  • get a number, name, date or quote wrong;
  • mistranslate, especially idioms, names and technical terms;
  • attribute a claim to the wrong person, or make a claim sound more certain than the article does;
  • fall out of date when the outlet updates its article.

Graphic-media flags can be wrong too: some upsetting media may not be blurred, and some harmless media may be.

The original article is always the reference. For anything important, especially health, money, legal matters, safety or elections, read the full article. AI summaries aren't advice.

Independent and neutral

SwipelyNews is independent. We aren't affiliated with any political party, government, religious or interest group, or with the outlets we summarise. The AI is instructed to summarise neutrally and to keep each claim attributed to its source. A summary reflects the outlet's reporting, not our opinion. AI can still introduce bias or errors, so please tell us if you notice any.

How we reduce mistakes

  • The summary is short and shows only the article's key facts. It's a pointer to the original, not a replacement for it.
  • The outlet is always credited, and the original article is always one tap away.
  • We keep a story's grouping with other outlets' versions (Full coverage), so you can compare them.
  • We review reports, and correct or remove stories that are wrong.
  • We work only with articles that are freely available, and we respect publishers' choices to opt out.

Report a wrong summary

In the App, tap ⋯ More › Report on the story and choose Misinformation (or Other). Add a short note on what's wrong if you can. You can also email support@swipelynews.com (subject: "Correction") with the story link, which you can get from Share. We review reports and correct or remove the story where needed.

Publishers can ask us to correct a summary or stop using their content at any time: see our Copyright & Publisher Policy.

Your data and the AI

We don't send your personal data to the AI. It only processes publishers' articles and media. Your feed is ranked by a separate system that doesn't use generative AI: see How your feed works.

Why we tell you this

We believe you should always know when text was written by AI. The EU Artificial Intelligence Act also requires AI-generated text published to inform the public to be disclosed (Article 50). That's why the AI summary label appears on every story and wherever our AI headlines appear, including notifications and shared links.

All language versions of this document are equally authoritative. If they differ, the interpretation most favourable to you applies.