Sentiment over generative AI is experiencing a course correction in the wake of Google and Microsoft’s missteps in implementing LLMs.
Publishers are now tackling the real-world implications of a tool that can spit out reams of copy in the blink of an eye for users with very little writing experience. Concerns are mounting over a flood of low-quality AI written stories swamping submission desks. Others, meanwhile, are asking serious questions about where AI is getting the data it’s repurposing.
Publisher Pain Point
INMA’s Peter Bale points out that the concern for publishers is two-fold when it comes to search-integrated LLMs — attributions and zero-click searches.Â
Publishers are up in arms at the prospect of generative AI aggregating knowledge on the web to create conversational answers that don’t cite where that information came from. As Bale points out, this raises some serious questions about “copyright, terms, accuracy, as well as the big question of payment”.
At the same time, there’s a real concern that search-integrated AI will exacerbate the issue of zero-click searches. This is where the user gets their answer from a Google Search feature — such as a featured snippet — without needing to click on a search result.
Wired has already reported about how the Bing chatbot was able to peek behind Wirecutter’s metered paywall and provide free access to the content. Wired noted that publishers were weighing up “whether to strike back at Microsoft”.Â
Traffic drives revenue. Yes, it’s more complicated than that, but not by much. Pick a monetization model and, at the end of the day, it either works or doesn’t based on how many people land on your site.
It was just over a year ago I wrote about how News Showcase was Google’s answer to the global regulatory push for both it and Facebook to pay to link to news articles.
The push in Europe has seen Google not only sign News Showcase agreements with publishers (which have been dogged by criticisms over the lack of transparency) but also launch a new automated licensing tool called the Extended News Previews (ENP) Program, which is now part of Google Search Console.
In the end, publishers may end up feeling like they need to strike back at search-integrated LLMs in order to protect their content from being cannibalized. Whether that’s through lobbying for regulatory change or lawsuits remains to be seen.
Given governments’ notorious struggle to stay abreast of tech changes, don’t be surprised if it takes years before we see any meaningful regulatory changes on this front., however.
AI-Walls
With this in mind, I asked State of Digital Publishing’s (SODP) Technical and News SEO Lead, Mahendra Choudhary, how he envisioned publishers might respond. Choudhary suggested that bigger news publishers — and at some point hosting providers/even ISPs — would start blocking AI bots from scanning their site.
Choudhary said: “No one wants their content repurposed by AI for free. But news publishers will either charge these AIs to crawl their data or prevent them from crawling their sites in the first place. This will make the AI results redundant and frustrating.”
He pointed out that Google has multiple bots for each segment of results — search, news, images, etc — and is likely using data from all of these bots for its Bard AI chatbot. ChatGPT, meanwhile, leans mostly on CCBot.
Choudhary said: “Eventually, all AI tools that scan the data will have to openly publish their bots names and the IP range they use to scan the web, similar to how Google and other search engines do. This will mean that web publishers can block them if they’d like to.”
While acknowledging the likelihood of CTRs for these websites experiencing something of a dip in traffic, he argued it was unlikely to make much of a dent. He noted that after the launch of features that drove zero click searches “publishers were still gaining clicks”.
