AI leaves web in the lurch

Driving the news: Tech circles were abuzz over the past week with news from the normally sedate browser world — the software category that has been shaping access to digital information since the '90s.

• Firefox last week debuted an experimental browser tool that provides AI summaries when users hover over links.
• Also last week, The Browser Company, maker of the Arc browser that's beloved by some power users, announced it was pivoting to focus on a new AI-powered browser called Dia.
• OpenAI has long been rumored to be working on its own browser, but has yet to ship anything.

Over the last two years Google, which customarily casts itself as champion of the open web, has steadily increased the prominence of its AI summaries in every aspect of search.

• At Google's I/O developer conference last year it announced the U.S.-wide rollout of the AI summaries, which sit on top of search results and allow users to get their answers without clicking through to source pages (while also sometimes providing made-up facts).
• At this year's I/O, the company said that AI Mode, which turns a user's search into an AI chat conversation, would now be a standard feature — although some early reviews have found its information unreliable.

Meanwhile, OpenAI made headlines with its announcement that it was purchasing Apple designer Jony Ive's AI device startup.

• Ive will now spearhead OpenAI's plan to sell new, non-smartphone gadgets that could bring generative AI answers more thoroughly into users' everyday lives.
• At Microsoft's recent Build developer conference the company introduced a new open project called NLweb, aimed at letting websites build their own chatbots to help site visitors access content.

What they're saying: "Increasingly, web pages — apps, articles, and files — will become tool calls with AI chat interfaces," Browser Company CEO Josh Miller wrote to explain why his firm was stopping further work on Arc.

Yes, but: As tech goes all in on rebuilding our web experiences with AI, there's no guarantee that the web will still be there when that job is done.

• With chatbots becoming users' default way to find out what's happening in the world, their makers pretend they can plaster this new interface layer over the internet without disrupting the data sources that feed it.
• But some media observers believe an AI-first web will choke off the money and attention that motivates web creators to keep extending the common knowledge pool.
• Many publishers are already seeing significant traffic and income declines from the shift toward AI search, though Google disputes there's a connection. And creative artists fear their work is being stolen or devalued.

This is everyone's problem. Of course the businesses and people that have built their work around the web are afraid — but AI makers should be worried, too.

• The web's vast treasury (and cesspool) of human creative work has accumulated since the 1990s because people wanted to share what they know either for financial or reputational gain, or just to advance a cause or do some good. That setup gave us everything from Wikipedia and YouTube tutorials to blogs and Reddit.
• Nearly all of the old-school web has already been fed into AI training databases for regurgitation by bots like ChatGPT. From now on, valuable new contributions are likely to sit behind subscription paywalls or depend on unsteady alternate means of support (membership programs, nonprofit grants, government funding).
• If AI undermines the incentives for human beings to update the web with their news, opinions and arguments, it will also cut off its own future.

The intrigue: Some in the web avant-garde are already anticipating a world in which the most ambitious or meaningful creative work takes place in what they're calling a "dark forest" web.

• They imagine creative communities that are purposefully isolated from the Silicon Valley bazaar, generating "anti-memes" and critical ideas without participating in social media's algorithmic competition or AI's sloppy reductivism.

The other side: AI firms have introduced modest efforts to feed money back to content providers. OpenAI, for instance, has made multiple deals with online publishers (including Axios).

• But it's hard to see how that kind of arrangement replaces the search traffic and ad revenue that's been the sustaining anchor for so many web publishers for the past decade.
• When it looked like Facebook's rise was threatening to hobble news publishers, the social media giant announced multiple programs to funnel them cash.
• But that support was fickle and fleeting, and site owners and creators know they can't sustain their businesses purely on tech largesse.

What's next: Advertising in AI chats is still in its infancy, but OpenAI has said it's going to build ads into ChatGPT, and its competitors won't be far behind.

• Everyone assumes this business will evolve rapidly, building on the performance-based model and personalization techniques that emerged in the search and social media eras.
• That can only further undermine the remnants of the web publishing industry, unless AI makers choose to share this new income with information providers.
• But they're spending billions on data centers, and their investors are expecting astronomical revenue growth, so no one should be surprised if they want to hold onto the lion's share.

The bottom line: Google shaped a search-based web on which independent publishers and individual contributors could survive, if not always thrive. Now AI is ready to turn that entire ecosystem into a legacy product.

Disclosure: Axios and OpenAI have a licensing and technology agreement that allows OpenAI to access part of Axios' story archives while helping fund the launch of Axios into four local cities and providing some AI tools. Axios has editorial independence.