Every side of every story, in your inbox by 7am.Subscribe
Theme
Appearance
Subscribe

About us

We read 105 outlets a day so you can see what each side left out.

attest.news is a daily news site. We read outlets from left to right, then show how each side told the same story and what each one left out. Below is exactly how we work, the rules we follow and where we fall short.

Real clippings from the November 16 edition: the same stories, headlined by the left and by the right. Each one opens its story.
105
outletsacross the US, World and India editions
87,866
headlines readsince October 18, 2026
336
stories writteneach one lists what it read
42
blindspots foundby code, from headline counts
01

From 1,117 headlines to 8 stories.

Four steps, run every day for every edition. Here is a real one: the US edition for Monday, November 16. Scroll to watch it move.

  1. We read every outlet, every day.

    For each outlet we follow, we run two Google News searches a day, one for politics and one for the economy. We keep every headline it published that day and drop exact repeats.

    On Monday, November 16, the US edition read 1,117 headlines from 48 outlets: 399 from the left, 387 from the center and 331 from the right.

    2
    searches per outlet, per day
    1,117
    headlines that day
  2. We group headlines into stories.

    Headlines about the same event become one story. We rank stories by how many different outlets covered them. Each outlet counts once, so a site that ran five headlines on a story still gets one vote.

    That day, 221 of the 1,117 headlines belonged to the 8 stories we wrote. The rest were smaller stories or fell outside politics and the economy.

    8
    stories that day
    33
    outlets on the top story
  3. Claude reads the reporting, then writes.

    For each story, Claude, the AI made by Anthropic, first reads 2 to 4 articles, ideally a wire report plus outlets from each side. Every fact must come from what it read, and every story lists those articles so you can check them.

    That day it read 35 articles to write 8 stories.

    2 to 4
    articles per story, by rule
    35
    articles read that day
  4. Code counts the coverage.

    The writer never touches the coverage numbers. Code counts how many outlets on each side ran each story, lists the ones that skipped it, and flags a blindspot when one side barely covered it.

    That day, 3 of the 8 stories were blindspots, all for the right.

    0
    coverage numbers set by the AI
    3
    blindspots that day
    See how a blindspot is decided
02

Who we read, and how we label them.

105 outlets across 3 editions, each with one of five labels. Left and lean left make up the left side. Lean right and right make up the right side.

Labels describe an outlet’s usual editorial stance and audience, never the quality of one article. US labels draw on public audience research, such as Pew Research Center’s studies of where Americans get their news, and on each outlet’s own opinion pages.

Left side20 outletsCenter11 outletsRight side17 outlets

Left

9
  • The Daily Beast
  • HuffPost
  • Mother Jones
  • MS NOW
  • The Nation
  • The New Republic
  • Raw Story
  • Slate
  • Vox

Lean left

11
  • ABC News
  • The Atlantic
  • CBS News
  • CNN
  • The Guardian
  • Los Angeles Times
  • NBC News
  • The New York Times
  • NPR
  • Politico
  • The Washington Post

Center

11
  • Associated Press
  • Axios
  • BBC
  • Bloomberg
  • CNBC
  • Forbes
  • The Hill
  • Newsweek
  • Reuters
  • USA Today
  • The Wall Street Journal

Lean right

8
  • Daily Mail
  • Fox Business
  • National Review
  • New York Post
  • Reason
  • Washington Examiner
  • Washington Free Beacon
  • The Washington Times

Right

9
  • Breitbart
  • The Daily Caller
  • The Daily Wire
  • The Federalist
  • Fox News
  • Newsmax
  • One America News
  • TheBlaze
  • Townhall
03

When a story counts as a blindspot.

A blindspot is a story one side barely covered. Code decides it from the headline counts, with three checks. The writer has no say.

Try a real story
US edition, October 24Justice Department sues to block Alaska Senate certification if Dan Sullivan loses

Outlets that ran it, of those we read that day

Left10 of 2050%
Center2 of 1118%
Right1 of 157%

The rate is the share of a side’s outlets that ran it. Center outlets count toward the total; the test compares left with right.

Right rate as a share of the left rate13%
0%25%, a quarter100%

The three checks

  • 6 or more outlets ran it13 outletsPasses
  • 3 or more on the left ran it10 on the leftPasses
  • The right at a quarter of the left’s rate, or less13%Passes
Blindspot for the right

Right outlets ran it at 13% of the left’s rate. That is a quarter or less. The story reached 13 outlets, and 10 on the left ran it. All three checks pass, so it is a blindspot for the right.

04

The rules Claude writes by.

Lines in quotes are copied word for word from the instructions the writer gets every day.

  1. Read the reporting first.

    Before writing, Claude reads 2 to 4 articles on the story, ideally one wire report (AP or Reuters) plus at least one outlet from each side.

    Get the facts, numbers, names and dates right.
  2. Every fact comes from a source.

    Nothing in a story may come from memory or guesswork.

    Every factual sentence must come from what you read. Never invent facts, quotes, numbers, names or events. If sources disagree, say so.
  3. Both sides get the same scrutiny.

    A story where the left buried a detail is written the same way as one where the right did.

    Be even-handed. Same scrutiny for both sides in every story.
  4. Headlines are quoted word for word.

    When a story shows how each side framed it, it names the real outlets and quotes their real headlines, in quotation marks.

  5. Context comes with citations.

    When a story turns on a claim about history, scripture, science, law or statistics, Claude adds the verified original with exact citations.

    Apply this the same way whichever side the claim helps or hurts. Never argue for a person or party.
  6. Titles stay plain.

    No teasers and no loaded words in the headline we write.

    Title: neutral, specific, under 90 characters, no clickbait.

What code does, so the writer can’t

These steps run after the draft is written. The numbers you see come from here.

  • Counts the outlets on each side that ran each story
  • Lists every outlet we read that skipped it
  • Flags blindspots with the rule above
  • Warns on drafts with no sources, very short stories or headline numbers that do not exist
  • Takes each photo from an article we read and credits that outlet
05

Where we fall short.

We would rather tell you than have you find out.

  1. We see what Google News indexed.

    We read each outlet through Google News searches. A story covered mainly by local papers, TV or podcasts can look smaller here than it was.

  2. Searches have a ceiling.

    Each search returns about 25 to 30 results, and we run two per outlet per day. On a very busy day we can miss some headlines from the busiest outlets.

  3. Headlines are not the whole article.

    Headlines are a strong signal of framing, but they are written by editors, not reporters.

  4. Every outlet gets one vote.

    A small website and a national paper count the same in our coverage numbers.

  5. Labels are our judgment.

    They describe an outlet’s usual stance and can lag when an outlet changes. They say nothing about whether a single article is accurate.

  6. The writer is an AI.

    Claude follows written rules and lists what it read, but it can still get something wrong. If you spot an error, write to hello@attest.news and we will correct it on the story page.

  7. In India we search in English.

    Hindi and other Indian-language reporting is not in the India counts.

06

Fair questions.

The ones we hear most, answered plainly.

Yes. Claude, an AI made by Anthropic, drafts every story after reading the reporting, under the written rules above. The coverage numbers are not written by the AI. Code counts them from the headlines.

You do not have to take our word for it. Every story lists the articles it was written from, every headline links to the outlet that ran it, and every coverage number is a count of outlets you can see on the story page.

We do, and every label is on this page. US labels draw on public audience research, such as Pew Research Center’s studies of where Americans get their news, and on each outlet’s own opinion pages. A label describes an outlet’s usual stance and audience, never whether one article is true. If you think we got one wrong, write to hello@attest.news and we will explain it or fix it.

Write to hello@attest.news with the story link. We check it against the reporting and correct the story page.

Because we read more outlets on the left than on the right: 20 against 17 in the US edition on November 16. Raw counts would make the right look quieter than it is. A rate asks a fairer question: what share of each side’s outlets ran the story?

Each day we take the biggest 6 to 8 stories, ranked by how many different outlets covered them, plus 1 or 2 blindspots when there are any. We skip lifestyle, celebrity, sports, weather, crime blotter and opinion-only pieces.

In Indian politics we use ‘left’ for outlets usually critical of the BJP-led central government and ‘right’ for outlets usually supportive of it. Labels describe usual stance and audience, never whether one article is accurate.

Check our work on today’s stories.

Every story shows its coverage counts, both sides’ real headlines and the articles it was written from.

Found a mistake? Write to hello@attest.news and we will correct the story page.