There Are Lies, Damn Lies, and Approval Numbers

Have you noticed the daily coverage of the Democratic Party’s approval rating? The breathless updates, the chyrons, the panel discussions about what it means for democracy? No? That is because there is none.

Have you noticed how the Democratic Party approval rating has been rising as Trump’s has been declining? No. Because there is no correlation.

Have you noticed the press’s frequent reporting on its own continuing twenty-year slide in public confidence? No. Because they are not reporting that either.

They present approval numbers when they support the narrative. Not when they don’t.

Fake News and the Art of Framing

Trump calls it fake news. The accusation lands, but the diagnosis is imprecise. Fact checking does not fix negative bias and framing. The press is not primarily in the business of fabricating facts. It is in the business of framing them. The most effective media manipulation does not require a single lie — it only requires a specific arrangement of truths. And it has become so systematic and teachable that artificial intelligence can now detect it in real time.

That is not an editorial problem. That is a methodology.

The techniques have names: Source Laundering, Strategic Burial, The Raising-Questions Construction, The False Concession, Attributed Motive. I catalogued them in an earlier piece — The Modern Journalist’s Field Manual — assembled entirely from observable practice.

No fabrication required. Just technique.

The Number They Don’t Run

The Democratic Party favorability numbers are this week’s example. The RealClearPolitics average as of late April 2026 has the Democratic Party at a net favorability of minus 23 points. Trump’s net approval over the same period: minus 18 points. The Democratic Party is less popular than Donald Trump.

Read that again. The party the press treats as the responsible adult alternative to Trumpian chaos polls worse than Trump himself.

Trump’s approval rating runs daily. Every half-point move gets a headline, an analyst, a chyron. The Democratic favorability collapse — historically deep, worse than the sitting president they are running against — gets no comparable treatment.

That is not an accident of editorial judgment. It is Strategic Burial applied at the category level — an entire story class disappears because it complicates the narrative rather than confirming it.

There is a third number that does not run daily. Gallup’s September 2025 poll put media trust at 28 percent — the first time it has ever fallen below 30 percent. Republican trust of the media is at 8 percent. Independents at 27 percent. Even Democrats, historically the most media-trusting group, are down to 51 percent. Pew found trust in national news organizations dropped 11 points between March and September 2025 alone.

The press decides which numbers become daily narratives. Presumably the press can be trusted to make that call honestly. Wink, wink.

The Polling Lever

There is one more thing the press is not telling you about those daily Trump approval numbers. They are statistically meaningless this far from an election and the press knows it.

A Berkeley Haas study found that polls conducted ten weeks before an election are accurate roughly half the time — a coin flip. The stated confidence intervals would need to be doubled even a week out to be statistically honest. Looking at the last twelve presidential elections, the eventual winner trailed in pre-convention polls seven times, with early polls missing the final margin by an average of nearly nineteen points.

The press employs statisticians. They have data editors. They know this literature. Running Trump’s approval as a daily news event six months from the midterms is not a lapse in methodological judgment. It is an editorial choice — the poll as headline, divorced from its predictive uselessness, deployed to sustain a narrative. When the numbers move against the story, the same polls that ran as breaking news quietly stop running. The methodology does not change. The editorial decision does.

The press claims these numbers reflect public sentiment, but sentiment is not the same as prediction — and the distinction is deliberately blurred. People are busy with their lives. An approval rating captures a reflex — the top-of-mind reaction to the last thing someone heard. Between elections, the news is something people observe but cannot control. When they vote, they have skin in the game — something with actual consequence. That is when opinions harden into decisions.

Flooding the Zone

The press is not just reporting. It is flooding the zone — a daily deluge of crafted narrative, framed polls, and attributed motives hitting people who have not thought deeply about any of it. This is not accidental. Political communication researchers have studied this for decades. The finding is consistent: media does not tell people what to think, but it tells them what to think about and how to weight it. Repeated exposure to a frame shifts perception even among people who believe they are not being influenced. The term is agenda setting. The mechanism is salience — whatever the press covers most becomes what feels most important, regardless of whether it is.

On one side of this transaction are professionals — journalists, editors, producers — who have spent careers mastering the techniques in the Field Manual. On the other side are ordinary people catching headlines between work and dinner. It is not a fair fight. The narrative arrives pre-assembled. The viewer absorbs it in the cognitive state of someone half-watching cable news, not someone deliberating carefully about a consequential decision. Then they run a poll. The poll does not measure public opinion. It measures the success of their own marketing. Then they report the poll as independent confirmation of the narrative.

The War as Case Study

If the press decides the war is bad, they flood the news with it. We are losing. Trump is looking for an offramp. We are out of ammo. Things were fine under Obama’s Iran deal and Trump wrecked it. Gas prices are high. Trump promised no more wars. Run those headlines for weeks, then poll the public on whether Trump is handling the war well. The result is not public opinion. It is the echo of their own coverage played back to them as data. Nobody has heard the other side in any balanced way because the other side does not get airtime. They see inflation at the gas pump and the grocery store and conflate it all. When people actually vote, that is when you find out what they decided after thinking about it.

George Stephanopoulos ran the Powerhouse Roundtable on This Week this morning arguing we were safer with Iran under Obama. The at least $50 billion in frozen assets Obama freed when he lifted sanctions under the JCPOA that were used to do evil, the proxy network, the October 7 funding, and the sunset provisions that lift all enrichment restrictions in 2030 — written into Obama’s own deal — buried. The frame requires you not to ask those questions, so the frame makes sure you don’t.

For the Powerhouse, they assembled a supporting cast of bobbleheads that included Chris Christie, for bipartisan credibility. He is apparently on a new diet where you only eat pie.

The Machine Notices

Take any major news article — pick one from the Times, the Post, the networks — and run it through a chatbot including my manual and a simple prompt: identify framing techniques, narrative choices, and bias indicators in this piece. The results are consistent and specific. The chatbot will name techniques from the Field Manual. It will also name techniques not in the manual — variants and hybrids it identifies independently. I did exactly this with a recent NYT Iran piece — The NYT’s Iran Story: Accurate Facts, Assembled Into a Brief — and the results were not subtle.

When a machine trained on no particular political agenda can reliably detect the methodology from the text alone, the methodology is real. It is not paranoia. It is pattern recognition. Systematic enough, consistent enough, and rule-governed enough to be learned, taught, and reverse-engineered by software.

This is what Trump is pointing at when he says fake news, even if he cannot articulate it precisely. The issue is not that reporters invent quotes or fabricate events. The issue is that the entire apparatus of story selection, source citation, headline writing, and question framing has been subordinated to narrative delivery.

The New York Times’ own audience is so enamored with its journalism now that more than 80 percent of its traffic goes to Wordle and the food app.

The public is developing herd immunity to the news media.