Title page 1

Contents 3

ACKNOWLEDGEMENTS 4

SUMMARY 5

1. INTRODUCTION: TRUSTED NEWS FOR AN INFORMATION EMERGENCY 8

THE INFORMATION EMERGENCY 8

NEWS SHOULD BE A SOLUTION, BUT STANDS ON SHAKY GROUND 8

2. AI COULD IMPROVE OUR NEWS LANDSCAPE 10

AI'S TRANSFORMATION OF NEWS IS UNDER WAY 10

THE RIGHT POLICIES COULD MAKE AI A FORCE FOR GOOD 11

THE FUTURE OF NEWS HANGS IN THE BALANCE 11

3. AI IS CURRENTLY AN UNRELIABLE INTERFACE FOR NEWS CONTENT 12

WHY AI SOURCING MATTERS 12

AI MAKES INVISIBLE EDITORIAL DECISIONS THAT SHAPE WHAT WE SEE 13

4. AI IS REDEFINING THE FINANCIAL INCENTIVES THAT SHAPE OUR NEWS ECOSYSTEM 20

ADVERTISING-BASED INCENTIVES ARE STARTING TO DISAPPEAR 20

AI LICENSING AGREEMENTS ONLY PARTIALLY REPLACE LOST REVENUE 21

NEWS WILL NEED TRANSFORMED BUSINESS MODELS TO SURVIVE 22

5. RECOMMENDATIONS 24

MAKE AI ANSWERS RELIABLE AND TRANSPARENT 24

SUPPORT THE COLLECTIVE LICENSING OF NEWS DATA FOR AI 26

CATALYSE INNOVATION IN THE NEWS ECOSYSTEM 28

6. CONCLUSION 30

REFERENCES 31

APPENDIX: AI ATTRIBUTIONS METHODOLOGY 35

DATASET GENERATION 35

SELECTING AI TOOLS 35

ANALYSIS 37

Tables 10

TABLE 1. Summary of the AI tools discussed in this report 10

Figures 14

FIGURE 3.1. AI tools predominantly draw on journalism in response to news queries 14

FIGURE 3.2. A narrow range of journalistic sources dominate in AI answers 15

FIGURE 3.3. AI creates new winners and losers 16

FIGURE 3.4. Representation of the BBC across AI tools varies 17

FIGURE 3.5. ChatGPT's top citations often align with OpenAI's publisher deals 18

FIGURE 5.1. Example of a transparency disclosure to precede an AI answer 25

FIGURE 5.2. Example of an AI interface to encourage more user choice over sourcing 25

Appendix Tables 36

TABLE A1. Summary of the AI tools discussed in this report 36