AI Overviews (AIO) are an existential threat to publishers. They steal their content and then, by creating derivative answers from it and presenting these to consumers without the need to click through, they steal the traffic that would otherwise generate the ad revenue that the publisher needs to survive. We’ve heard of 40%+ traffic declines from vulnerable publishers, and that’s before the technology has even been fully rolled out.
The question, however, is what they need to do about it. In the first instance, the immediate reaction is to block the crawlers that the AI companies use to appropriate their content. This approach is aligned to some regulator’s and publisher’s existing direction but on its own is ineffective.
Movement for an Open Web, recently filed a complaint about this topic to the CMA and the EC along with the Independent Publishers Alliance and the legal organisation, Foxglove. We call for an immediate injunction to protect publishers.
Blocking alone won’t stop the damage. The corpus of information which AI companies use to create their answers already exists and even without input from publishers they will continue to be able to give answers. This means that, blocked or not, Google and Apple will continue to siphon the traffic that publishers need to survive. By opting out of AI training the publisher loses any chance of getting the small amount of traffic that might arrive from the supplementary links within an AIO.
What is needed is a more holistic view of how AIO’s work and how publishers are rewarded.
In the first instance, we must introduce a payment system for harvested content in AI applications. Whether this is a one-off payment for harvesting or a per use payment when that content is used to enhance an answer, it is only fair that publishers receive just reward when their content is used to deliver a monetizable service.
Secondly, there must be fairness is how AIO’s are presented to consumers. Google has been sanctioned in the past for promoting its own services like maps, shopping, or travel in Search Engine Result Pages (SERPs). AIOs are no different. If AIOs are to be presented that presentation be on a fair and equal footing with other results which might answer the question as well or better.
Beyond this, there is the question of whether blocking or opt outs can even work on a technical basis. Businesses such as CDN provider Cloudflare are starting to offer publishers the ability to opt out of harvesting from AI companies other than Google and monetise their content using the Cloudflare network.
Adopting a proprietary and locked in Cloudflare solution is opening the door to a potential monopolistic gatekeeper. The last thing that the struggling publisher needs is yet another intermediary taking their revenues! Instead, the ability to opt in or out of harvesting to generate AI answers should be done through agreed, open and interoperable technical standards developed by the likes of the IETF.
We support publishers’ right to be compensated when it comes to AI harvesting and any use of their intellectual property – it is inconceivable that they should have no compensation if their content is stolen. We also believe that blocking is not effective. Publishers should have a fair chance of attracting reader and viewers through search results and when their content is used for an AI answer, they should be recompensed appropriately using terms they have genuine control over.