When AI reinvents the media business model

The media industry, when viewed through the lens of Artificial Intelligence, is both highly diverse and highly specific. It encompasses a wide range of categories : mainstream media, entertainment media, specialist publications, investigative journalism and opinion driven outlets. Each player fulfils a unique role, representing on average around 10 % of total employment, although this proportion varies significantly from one organisation to another.

An industry under pressure

The sector has already experienced several waves of rapid and profound transformation, particularly with the digitalisation of advertising formats. Overall, this shift has resulted in an average annual decline of approximately 2 % in net advertising revenues over the past decade, placing the industry under significant pressure : advertising revenue capture by digital players, the emergence of new programmatic intermediaries, shifting audience behaviours and more.

In this context, Artificial Intelligence may represent an additional layer of intensified competition, particularly through two major dynamics :

1. The commoditisation of competition, driven by the collapse in content production costs for virtually anyone, including both high quality and low quality content, as well as the proliferation of fake news
2. The emergence of new intermediaries further accelerating value cannibalisation, such as fully personalised daily media experiences generated by tools like OpenAI’s ChatGPT or equivalent platforms

At the same time, and based on our experience, Artificial Intelligence also creates significant opportunities to strengthen the position of media organisations, although unlocking this potential requires particularly careful strategic attention.

‍

- L’investigation, à travers des modèles d’analyse massive de contenus publics, tels des analyses de tendance sur les réseaux sociaux, et autres traitements massifs de contenus multi-médias, ou encore tels que l’enquête du Wall Street Journal sur les contenus Google Street View du New Jersey (pour y recenser les mauvaises pratiques environnementales des opérateurs télécoms de l’Etat concernant les abandons de cuivre)
- La vérification des faits ou la cartographie des points d’investigation, désormais possibles pour des coûts limités, sur des données publiques ou sur des dossiers journalistiques, en combinant différentes technologies permettant d’identifier les données quantitatives, éléments factuels ou croisements d’informations « posant question » donc à investiguer plus finement ;
- L’aide à la production, écrite ou pour tout média, encore utilisée de manière limitée à date, par exemple dans la presse écrite sur des suggestions de titres, mais en accélération sur des thèmes tels que les suggestions de montages, de synthèse écrites ou multimédias ;

2) De nouvelles interactions avec le lecteur 

Les médias ont désormais la connaissance de leurs auditeurs et lecteurs, les technologies disponibles et l’intérêt économique pour mettre en place une relation personnalisée et à valeur ajoutée avec leur audience.

Dans trois domaines notamment :
- La personnalisation des home-pages et interfaces, qui est généralisée sur la « vidéo à la demande », et a déjà été largement adoptée par les grands journaux aux Etats-Unis sur leurs apps et site web. Elle reste encore peu mise en place en Europe ; bien calibrée, les bénéfices sont importants, notamment à travers des taux d’opt-in (newsletters et notifications), de conversion des usages et au final de niveaux d’abonnement plus importants ;
- Des chabots « LLM », qui progressivement permettent, comme BILD Salut_ et Washington Post Climat, d’apporter une réponse structurée et sourcée sur les contenus et créer des formes d’interaction complémentaires qualitatives avec son audience ;
- L’élargissement des audiences à travers des formats étendus : différentes productions adjacentes à travers l’IA (traductions, versions audios, versions sous-titrées, etc…) sont désormais adoptées, apportant un élargissement de l’audience et des usages à coût très fortement réduit ; concernant des technologies moins matures (synthèses et teasers vidéos), des modèles assistés et contrôlés par l’humain apportent dès à présent de grands leviers de productivité ;

3. Advertising enhancement

AI also introduces innovative opportunities in advertising, such as :

• The revaluation of certain advertising inventories : current “brand safety” rules often penalise journalism, as blacklisting systems are primarily keyword based and fail to distinguish the actual context in which terms are used. For example, does an article mentioning “pure tech” create a positive or negative environment for an automotive advertisement, or even for a specific brand ? Similar issues also apply to video content
• The creation of multi local or personalised audio and video advertising campaigns at near zero marginal cost

Without aiming to be exhaustive, several major challenges emerge for media organisations seeking to embrace this transformation :

• Collaboration first : while these technologies and tools are now accessible at relatively low cost, relying exclusively on third party market solutions may require organisations to “give away their data”, whereas building fully dedicated systems independently may not always be economically viable. Multi stakeholder collaboration therefore becomes highly relevant. A few initiatives already exist, although they remain rare. One major project could, for example, involve the creation of a shared “ChatGPT for readers and audiences” jointly developed by French or European media players. In the field of audience measurement, media organisations previously succeeded in creating recognised collective structures such as Médiamétrie, demonstrating that this type of collaborative initiative is both possible and sustainable
• Quality and ethics : perhaps more than any other industry, media organisations operate under extremely high legal, professional and ethical standards. The response quality of current public LLM technologies still raises concerns for large scale media usage. However, these challenges are progressively being addressed through the development of “agents” and controlled intermediary layers capable of :
  • Moving closer to fully hallucination free responses
  • Preventing attempts to hack generative AI engines
  • Providing mechanisms for fact and data cross verification
• Long term vision and collective dynamics : major AI engines require media content to train their models. Unlike aggregators such as Google News, exhaustiveness is neither necessary nor even desirable for these technologies. In extreme cases, access to only a few archives and news streams may be sufficient, for example a right wing opinion newspaper, a left wing publication and a multi regional media source. To avoid a “prisoner’s dilemma” scenario in which the entire industry ultimately suffers, media organisations must adopt a collective, medium term and industrially coherent approach

There are also collective opportunities at the industry level, particularly regarding new measurement frameworks and the monitoring of best practices.

For example, through the work conducted by Eleven to develop the Observatoire des Médias sur l’Écologie, within a consortium bringing together Data for Good, Expertises Climat, Mediatree, Climat Médias and QuotaClimat, we provided audiovisual industry stakeholders with tools to quantify how they address the ecological crisis within their content.

To learn more about these topics and identify the impacts and opportunities relevant to your organisation, contact Jean-Charles Ferreri.

‍