Reach the people facts miss.

SHIFT is an EU-funded project testing how to reach the audiences disinformation targets first: the people fact-checks never get to.

Follow the first results See the problem
01/ 08 The problem SHIFT solves

Every dot is a person.

The bigger the dot, the more closely they follow the news.

Lines connect people who see and share each other's content.

02/ 08 The problem SHIFT solves

trust : engagement

X AXIS: how much people trust public institutions.

Y AXIS: how much they engage.

The coloured areas are population segments. The larger the area, the more people are in it.

03/ 08 The problem SHIFT solves

A fact-check goes out.

Platforms show it to the people most likely to engage with it.

That means people who already follow the news and trust its sources.

04/ 08 The problem SHIFT solves

It stays in the bubble.

Fact-checks mostly reach people who like facts.

The people most exposed to disinformation rarely see them.

The audience that needs it most sees it least.

05/ 08 The problem SHIFT solves

Everyone else drifts.

Disengaged audiences don't reject reliable information. It slips out of their feeds, then out of mind.

Over time, trusted sources move from top of mind to out of sight.

06/ 08 The problem SHIFT solves

That's who disinformation targets.

Those running disinformation campaigns aim at disengaged audiences on purpose.

Their content arrives through topics people already care about, and nothing is there to counter it.

07/ 08 The SHIFT solution

So we go through creators they already follow.

SHIFT works with creators inside the communities disengaged audiences live in: gaming, football, parenting, cost of living.

The message starts from what people care about, not from a rebuttal.

Every creator partnership is disclosed to the audience.

08/ 08 The SHIFT solution

Then we measure what moved.

We test what changes minds, not what gets likes. Pilots compare narratives across creators and countries to find what brings audiences back towards reliable information.

Moved Reached Not reached yet

This visualisation is illustrative. Results will be published as the pilots run.

How SHIFT works

01

Understand

We map who is drifting away from reliable information in 13 countries, what they care about, and which disinformation narratives reach them.

02

Design

With creators those audiences already follow, we build narratives that start from their interests rather than from a correction.

03

Test and measure

We pilot more than 40 interventions and measure what changes trust and resilience, not just reach.

See how we measure it
Method

How we measure what moved

We don't count likes. We compare each message with how a creator's content normally performs, then add that up across audience segments to see who is listening.

01

Start with a creator's normal.

Every dot is one of a creator's ordinary, non-political videos from the last 90 days.

Together they set the baseline: what a typical post from this creator looks like, and how much it usually varies.

At least 8 organic posts are needed. One-off viral hits are trimmed so they don't skew the baseline, and a creator who posts on several platforms gets a separate baseline on each.

02

Each video gets one engagement score.

Reach, watch-through, likes, comments and shares are combined into a single number.

The weights are a strategic choice. To build trust with hard-to-reach audiences, reach and watch-through count most.

03

Then a political post is measured against it.

We ask one question: did this message do better or worse than this creator's normal content?

The answer is the lift, measured in standard deviations. Zero means it performed like any other post.

04

Every creator gets their own yardstick.

A gamer on TikTok and a parent on Instagram can't be compared on raw numbers.

Scoring each post against its own creator's baseline puts every platform, niche and audience size on the same scale.

05

Each audience is a mix of segments.

We estimate who each creator's followers are, using their interests, cultural markers and past performance.

The mix is fixed before a campaign starts, so results can't shape the map they're measured on.

06

Then each segment is classed.

Above the baseline is an echo chamber: an audience that already agrees. Below the floor is unreachable. Everything in between is reachable.

Reachable is the target. Political content will always do worse than a creator's normal posts. Reachable audiences still watch it, without already agreeing.

07

Results are pooled by segment.

Each post's lift counts towards every segment in its creator's audience, in proportion to the mix.

Bigger creators count more, but only by the square root of their following, so many small creators aren't drowned out by one large one.

The bars show the range we're 95% confident in, found by resampling creators.

08

Then we find which message lifts reachable segments.

Each reachable segment's results are compared across message versions. Lift is how much better the strongest message does than that segment's average political post.

Echo chamber and unreachable segments are greyed out. They're not what SHIFT optimises for.

09

Lift builds over time.

Reachable audiences rarely respond the first time. After each round, the strongest messages are refined and tested again.

In this example, the drop for Disengaged narrows from −0.12 to −0.04 in the second round, and the lift grows with it.

Illustrative data.

What the numbers can't tell us

Engagement isn't persuasion

High lift can mean agreement or outrage. Over time we add comment sentiment and real actions, like sign-ups, wherever they can be measured.

Audiences, never individuals

Segments are estimated at audience level only. No individual follower is ever profiled for their political views.

Labelled and disclosed

Creator content is labelled under EU political advertising rules, and we track whether labelling changes how people engage.

Where the project is now

Done so far

  • 35expert interviews across 13 countries
  • 5countries with focus groups
  • 13country reports on national disinformation landscapes
  • ~4,000mis- and disinformation narratives classified into typologies
  • 13countries surveyed about 1,000 adults each, on how and why people respond to disinformation

Coming next

  • A cross-country dataset and comparative report on the impact of disinformation
  • National insight briefs for all 13 countries
  • 40+ tested pilot interventions and a proven narrative strategy methodology
  • An open-access narrative toolkit in five or more languages

Nine partners, twelve EU countries and Moldova

SHIFT is led by the Center for the Study of Democracy and brings together researchers, civil society organisations and communications specialists.

Center for the Study of Democracy GLOBSEC European Movement Ireland Make.org Logiq Media Alliance4Europe CeMAS Funky Citizens Rīga Stradiņš University

Working on disinformation?

Who it's for

People countering disinformation in civil society, policy, media and research. If you want to reach beyond the audiences who already trust you, SHIFT's results are for you.

Follow the first results

We'll share findings, country briefs and the narrative toolkit as they're ready. Questions? Get in touch with the team.

Email the team