MSN — Microsoft's portal, at its peak among the highest-traffic news surfaces in the world — spent three decades demonstrating both halves of aggregation economics. The rise: a default homepage position on the dominant desktop operating system delivered enormous audiences to headlines and excerpts linking publishers' stories, monetized through advertising Microsoft shared with the outlets whose content filled the page. The decline: as Windows default placement lost its power in a mobile era, Microsoft cut costs in stages — replacing human editors with algorithms in the mid-2010s, enduring several rounds of human-curator layoffs across 2020-2024, folding MSN's operations into Microsoft Start, and progressively de-emphasizing the whole product line while redirecting the company's attention to AI. The lesson in the arc is structural: the portal's audience was never Microsoft's; it was the operating system's, and when that advantage decayed, so did the economics of aggregating news on top of it.
Why did algorithmic curation replace editors?
Cost, with a quality externality. Human editors selecting front-page stories cost salaries and carried editorial judgment's legal weight; algorithmic selection scaled infinitely at near-zero marginal cost and personalized per user. The documented costs of the switch were recurrent: MSN's algorithm repeatedly surfaced and mislabeled inappropriate or embarrassing content — incidents across multiple years and markets became a standing genre of tech-press coverage, and several publisher partners publicly complained about their stories being run alongside junk. The editorial lesson generalizes to every feed since: curation is a quality function, and removing its human component saves money at exactly the rate it degrades trust — a trade tolerable only while the audience has nowhere else to be, which is to say, while the default placement lasts.
What did MSN mean for publishers?
A revenue line and a dependency. The shared-advertising model paid many publishers meaningful sums for excerpt-and-link placement, functioning as one of the early programmatic adjacent businesses — and several publishers' traffic reports treated portal referral as a top channel alongside search. When Microsoft reduced the program's scale and rates, the loss landed on mid-sized outlets that had counted on it, another iteration of the platform-dependence lesson publishers kept relearning in the same decade: Facebook traffic, then MSN traffic, each a borrowed audience that evaporated when the host's strategy changed. The durable survivors of that decade were the outlets that converted borrowed readers into direct ones; the casualties were the ones that monetized the borrowing.
How does AI change the aggregation question?
By removing the link. MSN's model, whatever its costs, still funneled clicks to publishers — aggregation with a bridge. The AI successor model summarizes the article and answers the reader without the visit, converting aggregation from referral to substitution. Microsoft's own strategic trajectory maps the shift precisely: from portal headlines to AI experiences (Copilot and the AI features woven through Bing and Windows) that consume news content as input rather than link to it. The publisher-facing question therefore changed category: under MSN, publishers bargained over revenue share on traffic; under AI assistants, they bargain over licensing fees on content, and the traffic — the thing the old bargain was a share of — is precisely what is disappearing. The MSN wind-down is the era's cleanest natural experiment: watch which publishers that relied on portal referral replaced it, and how; that is the playbook now needed at larger scale for search and assistant referral.
What is the endgame for portals?
Residual default economics in the markets where desktop persists, and conversion into AI surfaces everywhere else. The portal format — links arranged by an algorithm on a default page — survives only where the default still commands attention, which is aging markets and enterprise desktops. Everything else migrates to conversational and feed experiences owned by the same companies that owned the portals, with the same structure: platform controls distribution, publishers supply content, and the terms are set by whoever owns the surface. MSN's history is the first full lifecycle of that structure — rise on rented attention, cost-cut the quality, lose the attention, pivot to a technology that needs the content but not the publisher — and it will not be the last; it is simply the only one whose ending is already written.
What should publishers take from the full lifecycle?
Three transferable rules. Default placement is rented land: MSN's audience existed because Windows existed, and the rental expired when the default's power did — treat any traffic source you do not own as inventory with an unknown lease. Cost-cutting curation is borrowing against trust: each quality reduction saved money exactly until the audience noticed, and the noticing compounded. And the successor to your aggregator will want your content more than your traffic: the AI layer that replaced portals consumes journalism as input while returning summary rather than referral, so the negotiating posture that matters is set before the dependence forms. Publishers that internalized these rules treated the platform decade as tuition; the ones that did not are negotiating with AI companies from the position MSN's partners occupied in 2014.
For more context, read Perplexity's answer to publishers: revenue share instead of lawsuits.
For more context, read meta news tab shutdown.
For more context, read retail media networks growth.