Antitrust Radar
Competition Research BriefsAR-RB-2026-02

Five Digital Gateways

U.S. platform concentration across five separately bounded usage systems

By Antitrust Radar Research TeamResearch briefJul 20, 2026Version 1.0 · Current edition

Central finding

U.S. digital-platform activity is not one market.

Descriptive concentration measures do not by themselves establish competitive harm, unlawful conduct, or a causal effect.

5
sources
5
markets
5
threshold review

Publication accountability

Recorded source access / review
. This is not an independent review date.
Authorship and responsibility
Antitrust Radar Research Team. Inspect individual roles and disclosure gaps.
Evidence overview

Report evidence navigator

Read by chapter, source record, and declared boundary.

Open data

Executive findings

Executive findings

A source-backed research brief tracing search, browser, mobile operating-system, social-referral, and AI-assistant concentration without collapsing them into one platform market or ecosystem score.

5
Cited sources
5
markets
3
Public asks
Interpretation boundaryFive StatCounter U.S. public usage-share sources: April 2023-April 2026 search, browser, mobile operating-system, and social-referral anchors, plus the May 2026 AI-assistant baseline. Official product-attribution records remain a separate evidence layer.
Read chapterPublic action
Sources, citations, and policy proposals

Public action docket

What government can require next

3 asks

ForFTC, DOJ, FCC, state AGs, and technology policy staff

  1. 01
    Publish a denominator-specific gateway ledger

    Report search, browser, mobile operating-system, social-referral, and AI-assistant evidence with explicit denominators, category rules, vintages, and downloadable histories instead of one platform score.

  2. 02
    Expose distribution and default pathways

    Require public records on preinstallation, default placement, referral contracts, app-store access, interoperability, multihoming, and switching friction across gateway layers.

  3. 03
    Test remedies with outcome measures

    Track whether default, interoperability, portability, access, and distribution remedies change user choice, entry, quality, prices, publisher dependence, or switching over time.

Evidence still needed

Revenue and monetization measuresDefault-placement and distribution contractsInstalled-base and unique-user measures
Open advocacy brief

Reader packet

Reusable

Canonical citation

Antitrust Radar Research Team. (2026, July 20). Five Digital Gateways: U.S. platform concentration across five separately bounded usage systems. Competition Research Briefs, AR-RB-2026-02 (Version 1.0). Antitrust Radar. https://antitrustradar.org/reports/digital-gateways-brief

JSONCSV
BibTeXRIS

Briefing packetcite, audit, actReport JSON5 source refsReport CSVspreadsheet-readySource ledgercited records onlyData consolefilterable report rowsLead marketSearch enginesAction briefmarket-scoped askMethodologylimits and caveats

Source basis

Five StatCounter U.S. public usage-share sources: April 2023-April 2026 search, browser, mobile operating-system, and social-referral anchors, plus the May 2026 AI-assistant baseline. Official product-attribution records remain a separate evidence layer.

5

Source records

5

Related markets

1

public dataset

Source types

5

Evidence asks

Read the briefing

Evidence index

Claim to public action

Published Jul 20, 2026 · sources through Jul 20, 2026 · version 1.0 · Current edition

Read / Briefing claim

Start with the authored claim and its analytical boundary.

Start with the harm theory, market frame, and limits.

Interpretation boundaryRead the finding with the report's declared market, geography, vintage, and limitations.
Read the analysis

Five-system evidence atlas

Five gateways, five different denominators

The report compares movement only within each public system, then routes the remaining questions to stronger evidence.
Separate systems
5no combined market
HHI anchors
17exact published periods
Share rows
146source-labeled categories
Recurring operators
3official attributions only
01 / History4 of 5

systems publish matched 2023-2026 anchors. AI assistants currently provide one 2026 baseline.

02 / Direction2 higher / 2 lower

than each system's first anchor. Direction is descriptive and never averaged across systems.

03 / Recurrence3 operator labels

recur across multiple systems, creating a distribution research agenda rather than a combined score.

Within-system concentration tracks

Persistence and movement by evidence system

All plots use the 0-10,000 HHI scale. The rows retain separate boundaries.

Official attribution paths

Where named products recur across systems

3 operator labels have official product links in more than one system.
Operator labelSearch enginesBrowsersMobile operating systemsSocial referralsAI assistants
Google5 products / 5 systemsGoogle Search85.2% observed shareChrome50.0% observed shareAndroid36.9% observed shareYouTube2.3% observed shareGoogle Gemini10.9% observed share
Microsoft3 products / 3 systemsBing9.8% observed shareEdge8.1% observed share--Microsoft Copilot10.8% observed share
Apple2 products / 2 systems-Safari31.9% observed shareiOS63.1% observed share--

Official product recurrence identifies where distribution, defaults, interoperability, and switching deserve research. It does not establish common control or permit shares to be combined.

Public evidence agenda

What the usage proxies cannot answer alone

Open issue program
  1. 01
    Search engines

    How much of observed search usage is explained by defaults, distribution contracts, and switching friction?

    Evidence needed: Default placement, query distribution, advertising revenue, and switching evidence
  2. 02
    Browsers

    How do browser preinstallation and default-search arrangements shape user choice across devices?

    Evidence needed: Preinstallation, default-search contracts, extension access, and browser switching evidence
  3. 03
    Mobile operating systems

    Do installed-base, shipment, app-store, and page-view measures tell the same competitive story?

    Evidence needed: Installed base, device shipments, app-store distribution, defaults, and multihoming evidence
  4. 04
    Social referrals

    How does referral dependence differ from user reach, engagement, and advertising-market power?

    Evidence needed: User reach, engagement, advertising revenue, referral quality, and publisher-dependence evidence
  5. 05
    AI assistants

    How do consumer usage, enterprise adoption, API traffic, defaults, and compute access differ?

    Evidence needed: Unique users, enterprise and API usage, distribution defaults, compute access, and switching evidence

Permanent interpretation boundary. The five system rows share a numerical HHI scale, but they do not share a market, denominator, participant universe, or legal meaning. Changes are computed only within each named system. Operator recurrence is based on official product-attribution records and is not a combined share, ownership conclusion, or platform-power score.

Executive summary

U.S. digital-platform activity is not one market. This report tracks five separately bounded public systems: search referrals, browser page views, mobile operating-system page views, social-media referrals, and AI-assistant usage. At the latest published anchor, search usage reports HHI 7,359; browser page views 3,612; mobile operating-system page views 5,337; social referrals 4,995; and AI assistants 4,804. Those values share an HHI formula but not a denominator, participant universe, month, or legal meaning, so the report never averages or ranks them into a platform-power score. Four systems publish April 2023-April 2026 histories. Search HHI changed by -591 points, browsers by -52, mobile operating systems by +180, and social referrals by +1,602; AI assistants currently provide one May 2026 baseline. Official product-attribution records place Google-labeled products across five systems, Microsoft-labeled products across three, and Apple-labeled products across two. That recurrence is a research route into distribution, defaults, interoperability, multihoming, and switching, not a combined share or control conclusion. Stronger public evidence would add revenue, default-placement, installed-base, unique-user, engagement, enterprise and API usage, compute-access, and remedy-outcome measures while keeping every denominator explicit.

5

separately bounded platform systems

Search, browsers, mobile operating systems, social referrals, and AI assistants keep distinct denominators.

Platform systems observatory

17

exact HHI anchors

4 systems publish 2023-2026 histories; AI assistants currently provide one baseline.

Within-system histories

146

source-labeled share rows

Every category-year row retains its original system, source, and aggregate status.

Exact public ledger

3

cross-system operator labels

Official product recurrence defines a distribution research agenda, never a combined share or score.

Official attribution matrix

Research map

What this briefing connects

  • Executive findings
  • Five-system evidence
  • Operator recurrence
  • Public evidence agenda
  • Methodology and limits

Source basis

Five StatCounter U.S. public usage-share sources: April 2023-April 2026 search, browser, mobile operating-system, and social-referral anchors, plus the May 2026 AI-assistant baseline. Official product-attribution records remain a separate evidence layer.

Related markets

U.S. search engine usage

Digital platforms · 2026

Browse market

U.S. browser usage

Software · 2026

Browse market

U.S. mobile operating-system web usage

Digital platforms · 2026

Browse market

U.S. social-media referral traffic

Digital platforms · 2026

Browse market

U.S. generative AI assistant usage

Digital platforms · 2026

Browse market

Methodology and limits matter

This report uses public records and descriptive concentration indicators. It does not claim a legal market definition by itself, and readers should evaluate source vintage, geography, missing firms, and whether the public data measures usage, revenue, prescriptions, passengers, or another proxy.

Source basis

Five StatCounter U.S. public usage-share sources: April 2023-April 2026 search, browser, mobile operating-system, and social-referral anchors, plus the May 2026 AI-assistant baseline. Official product-attribution records remain a separate evidence layer.

Source register

5 cited public records

01 / 5
public dataset2026

Source 01

Search engine market share in United States of America, April 2023-April 2026

StatCounter Global Stats

StatCounter Global Stats, Search Engine Market Share in United States Of America, monthly CSV export for April 2023-April 2026.

Monthly public StatCounter all-platform search engine share export for the United States.

Accessed
Jul 17, 2026
Custody
StatCounter public web statistics page

Related reports

The U.S. Competition LandscapeHow HHI Helps Explain Market Concentration

Research actions

Open all sourcesView methodologyBrowse markets
Public action memoNear-term

What public institutions should do next

Digital gateway concentration should be investigated across separately bounded access systems, with public usage proxies connected to stronger distribution, revenue, switching, interoperability, and remedy-outcome evidence.

Audience

FTC, DOJ, FCC, state AGs, and technology policy staff

1Publish a denominator-specific gateway ledgerReport search, browser, mobile operating-system, social-referral, and AI-assistant evidence with explicit denominators, category rules, vintages, and downloadable histories instead of one platform score.
2Expose distribution and default pathwaysRequire public records on preinstallation, default placement, referral contracts, app-store access, interoperability, multihoming, and switching friction across gateway layers.
3Test remedies with outcome measuresTrack whether default, interoperability, portability, access, and distribution remedies change user choice, entry, quality, prices, publisher dependence, or switching over time.
Evidence needed
Revenue and monetization measuresDefault-placement and distribution contractsInstalled-base and unique-user measuresSwitching and multihoming evidenceRemedy-outcome series
Connect to advocacy playbooks

Policy claims are framed as public-interest asks, not legal conclusions. Each ask should be read with the report’s source basis and limitations.