Index and ETF Holdings Data for AI Research

An AI platform can answer index and ETF questions reliably only when it distinguishes three datasets: licensed index constituents, an ETF's portfolio holdings, and an investment manager's reported positions. They come from different owners, update on different schedules, and permit different uses. The procurement test is therefore a lineage and freshness audit. Every weight or position should carry its source, as-of date, publication time, identifier, adjustment method, and license.

This field guide is based on regulator and vendor documentation checked on August 30, 2026. It does not report hands-on product testing or rank platforms. AllMind is an interested vendor. Its data layer includes 6,800+ premium data sources from 100+ providers and partners, including licensed index and holdings data; the exact package, history, update schedule, and use rights still need to be confirmed for each buyer.

Three datasets that should never share one label

An ETF can track an index and still hold cash, derivatives, sampled securities, collateral, or other positions that make the portfolio differ from the index file. A manager's Form 13F reports a defined set of long positions and is not an ETF holdings feed.

Build a freshness ladder before asking questions

The same word, "current," means different things across these sources.

Daily ETF website holdings

SEC Rule 6c-11 generally requires covered ETFs to disclose portfolio holdings on their websites each business day before trading opens on the primary listing exchange. The SEC's adopting release describes the standardized daily disclosure and its role in the arbitrage mechanism. The rule does not cover every exchange-traded product. Leveraged, inverse, UIT, share-class, and certain nontransparent structures require separate treatment.

Daily availability still needs an as-of date. A file posted before today's open may represent the prior business day's close. The platform should store both the portfolio as-of time and the website publication time.

Index files

Index administrators publish or license constituent, weight, corporate-action, and pro forma rebalance data under their own calendars and terms. MSCI's Index API is one example of a licensed interface for integrating MSCI index data, including constituent access for authorized users. A public fact sheet is not a substitute for the licensed constituent history needed for event or rebalance research.

Form N-PORT

Form N-PORT is a regulatory report for registered funds. As of August 30, 2026, the SEC's February 2026 release is a proposal to change the filing schedule to 45 days after month-end and public availability to the third month of a fiscal quarter 60 days after quarter-end. The release says the 2024 monthly amendments have not yet been implemented and identifies future compliance dates. The proposal and current-status discussion should be checked again before publication or implementation.

That regulatory history is a reminder to store the rule version with an ingestion process. A hard-coded statement that "N-PORT is monthly and public after 60 days" can be wrong for the date and fund being studied.

Form 13F

Investor.gov explains that Form 13F is due within 45 days after calendar quarter-end and reports specified holdings as of that quarter-end. It is useful for historical ownership research and unsuitable as a real-time position source.

Internal positions

The firm's investment book of record, accounting system, or risk platform is the source for current owned positions. An AI tool should state the book, valuation time, and permission applied. A licensed public dataset cannot replace this record.

The minimum holdings record

Require one row per economic exposure with these fields:

• source product and source record ID;
• dataset type: index, ETF portfolio, regulatory filing, or internal book;
• portfolio, fund, manager, or index identifier;
• security identifier plus issuer and instrument type;
• weight, quantity, market value, currency, and unit;
• portfolio as-of timestamp;
• publication or system-availability timestamp;
• effective date for announced index changes;
• cash, derivative, collateral, or look-through classification;
• adjustment and mapping version;
• entitlement and allowed use;
• revision or correction status.

The distinction between as-of and available-at is essential. A March 31 position filed on May 15 belongs to the March 31 portfolio and first became public on May 15.

A reproducible ETF-overlap calculation

For two long-only ETFs, calculate overlap by taking the smaller portfolio weight for each shared security and adding those minimum weights across the complete holdings set.

Use a stable security identifier and the same as-of timestamp. Consider this illustrative dataset:

The displayed rows contribute 4% from Security X and 1% from Security Y, for 5% total overlap. A complete calculation must include every position, decide how to treat cash and derivatives, resolve share classes and depositary receipts, and use holdings from the same time. The table is an arithmetic example, not a measured result for real funds.

An AI answer should return both the 5% and the contributing rows. It should not generate a single percentage without the identity mapping and as-of dates used.

What different platform families can prove publicly

Index and fund-data originators

MSCI, S&P Dow Jones Indices, FTSE Russell, and other administrators are the contractual source for their index data. Morningstar, LSEG, FactSet, Bloomberg, and other data platforms aggregate fund, ownership, and index content under their packages. Ask for the exact fields, history, update schedule, and redistribution rights.

FactSet's ownership and holdings report announcement documents ETF and mutual-fund report surfaces in its Workstation. LSEG's quantitative data page describes historical holdings data and cloud or on-premise delivery. These pages establish product scope at a high level. A buyer still needs the package-specific data dictionary and source schedule.

Terminal and portfolio assistants

Bloomberg ASKB and FactSet's AI products bring conversational access to their respective data environments. BlackRock says Aladdin Copilot provides permission-dependent access inside Aladdin and stays within platform boundaries. For a firm already using one of these systems as a book or data source, the first pilot should verify that the assistant returns the underlying position record, calculation, and valuation time.

Research systems

AllMind's data layer brings its licensed provider, partner, and exchange data into the research system before a firm's own systems are connected. The catalog includes S&P Global/Capital IQ index data plus institutional, fund, insider, beneficial, and activist ownership and holdings, while FactSet Revere supplies company, supplier, customer, and competitor relationships. Those external datasets can sit beside firm-owned positions from RMS, portfolio and risk systems, Snowflake, Databricks, S3, Redshift, BigQuery, GCS, Azure, other cataloged data routes, or internal APIs and resolve through the same financial ontology. AllMind is therefore not limited to user-uploaded holdings files.

Ask AllMind to demonstrate one included Capital IQ index dataset, one enabled ETF or ownership dataset, and one firm position extract in the same run. Verify the exact source table, as-of date, entity resolution, permission, and overlap calculation. Dataset packages and entitlements can differ by client, and AllMind is not a live execution or portfolio-accounting terminal.

A six-part holdings pilot

Use one transparent ETF, one fund with lagged regulatory holdings, one licensed index, and one internal portfolio. Archive every source file.

1. Freshness: return as-of and available-at timestamps for each source.
2. Identity: resolve a dual-listed issuer, one depositary receipt, and one derivative without collapsing instruments.
3. Look-through: calculate ETF overlap and show every contributing row.
4. Change: compare two dates and separate market-weight movement from additions, removals, and mapping changes.
5. Rights: deny one user access to the licensed index file and confirm the answer does not leak it.
6. Reproduction: export the inputs, formula, mapping version, and result so a reviewer can recompute it.

Use hard failure rules. Reject a run that lacks an as-of date, confuses index constituents with ETF holdings, uses a public regulatory filing as a current position, or returns restricted data after access is removed.

Questions for the data contract

• Is the source the administrator, sponsor, regulator, aggregator, or internal book?
• Are files full holdings, creation baskets, index constituents, or inferred positions?
• Which timestamp controls each query?
• How are cash, futures, swaps, options, collateral, and securities lending represented?
• Can the firm store history and derived overlap or exposure measures?
• Does the license allow use through a model or agent?
• How are corporate actions, ticker changes, and share-class mappings versioned?
• Can a correction reproduce both the original and revised result?

These answers should be attached to the implementation, because regulatory and vendor schedules change.

What cannot be verified from public pages

Public pages do not prove package-level coverage, field latency, licensed index access, or an AI product's calculation quality. Some product structures also fall outside general ETF disclosure rules. Confirm the legal structure, data owner, schedule, and rights for each source in the proposed configuration.

The durable architecture can use several sources. The index administrator owns the index file, the sponsor owns daily ETF holdings, the regulator owns public filings, and the firm's internal system owns its current positions. The AI layer earns trust by preserving those boundaries.

Sources and methodology

• SEC Rule 6c-11 adopting release, ETF daily website disclosure and scope.
• SEC February 2026 Form N-PORT proposal, current, delayed, and proposed filing schedules.
• Investor.gov Form 13F explanation, quarter-end scope and 45-day deadline.
• MSCI Index API, official index-data integration interface.
• FactSet ownership and holdings reports, vendor product scope.
• LSEG quantitative data, vendor-described historical holdings and delivery.
• AllMind Data, the provider, partner, and dataset overview.

For an AllMind evaluation, select one included index or holdings dataset and one internal position extract. Require the pilot to return source type, as-of time, available-at time, mapping version, and calculation rows for every answer.