AI disclosure in SEC filings: how many 10-Ks and 10-Qs mention AI, counted from EDGAR, 2019 to 2026

In 2025, 3,271 of the 6,431 10-K annual reports filed on EDGAR, 50.9%, contained the phrase “artificial intelligence”, up from 432 of 6,660 (6.5%) in 2019.¹⁴ In 2026 to 2026-11-10 the count is 3,850 of 6,183 10-Ks filed, 62.3%, and 1,206 of them (19.5%) also use the phrase “generative AI”, a term that appeared in two 10-Ks in all of 2022.¹²⁴

Those are counts of filings, not of companies’ views. A 10-K counts once whether it names artificial intelligence in one boilerplate risk factor or across forty pages of a business description, and the counts say nothing about what a company does with AI in its own finance work. This site’s filings log does that part by hand.

The question and the rule for counting

The question people put to search engines is “how many companies mention AI in their 10-K.” The answer here is a count of 10-K filings, which is close to a count of companies because a company files one 10-K a year, though not equal to it: an amended 10-K/A is not counted, and a company that changes its fiscal year can file two 10-Ks in one calendar year.

The rule is simple and strict. A filing counts when the full text of any document in it contains the exact phrase. “AI” alone is not searched, because the two letters also match tickers, company names and other words, so a filing that only ever writes “AI” is not in these counts. The phrase “generative AI” is counted separately, and so is “machine learning” in the table at the foot.

Method: queries, windows, denominators and the 10,000-hit cap

EDGAR full-text search returns at most 10,000 hits for a query, so no query here covers more than one form in one calendar quarter of filing dates.¹ Every page of results for every quarter from 2019-Q1 to 2026-Q4 was read, and the hits, which are documents, were reduced to distinct accession numbers, which are filings. The largest 10-K quarter was 2026-Q1, with 3,261 matching documents in 3,216 filings.¹

The denominators are EDGAR’s quarterly form indexes, which list every filing by form type.⁴ Each index was read and its 10-K and 10-Q accession numbers counted, amendments excluded. Over the eight calendar years the 10-K count runs from 6,183 (2026, to 2026-11-10) to 7,538 (2022).⁴

The queries, run on 2026-11-10, were the exact phrases “artificial intelligence” and “generative AI” for forms 10-K and 10-Q, with the filer’s SIC code kept for 10-Ks filed in 2025.¹²³ The scripts that ran them and computed every figure on this page are in the site’s repository under scripts/cfonews-ai/articles, and the counts they produced are committed beside the article.

Share of 10-Ks mentioning AI, by year

The rise has two steps. From 2019 to 2023 the share climbed by two to six points a year: 6.5%, 8.3%, 12.1%, 15.1% and 18.0%.¹⁴ Then it doubled in a year, to 36.4% of 10-Ks filed in 2024, and passed half in 2025.¹⁴

The first-quarter numbers carry the most weight, because most companies have a December year-end and file their 10-K between January and March. In 2025-Q1, 2,560 of 5,007 10-Ks named artificial intelligence; in 2026-Q1, 3,216 of 5,132.¹⁴

The 2024 jump, in the first quarter

First quarters isolate the December year-end filers, the bulk of 10-K filers, and show where the jump sat. In 2023-Q1, 851 of the quarter’s 10-Ks named artificial intelligence, 15.5%. In 2024-Q1 the count was 1,713, or 36.1%: the share more than doubled between two filing seasons twelve months apart.¹⁴ It rose again to 51.1% in 2025-Q1 (2,560 filings) and 62.7% in 2026-Q1 (3,216).¹⁴

The generative AI count moved in the same season, from 20 10-Ks in 2023-Q1 to 365 in 2024-Q1.² The count cannot say why companies added the words; the passages themselves can. In this site’s filings log, 27 of the 41 10-K entries were filed in 2026-Q1, and each quotes what the company wrote.

Generative AI: when the term enters filings

“Generative AI” is close to absent before 2023. Full-text search finds it in one 10-K filed in 2021 and two in 2022.² It reached 69 10-Ks in 2023, then 511 in 2024 and 870 in 2025.² In 2026 to 2026-11-10, 1,206 10-Ks use the phrase, about one in three of those that name artificial intelligence at all.¹² Measured that way, as a share of the 10-Ks that name artificial intelligence, generative AI went from 5.4% in 2023 (69 of 1,283) to 21.2% in 2024, 26.6% in 2025 and 31.3% in 2026 to date.¹²

The same count for 10-Qs

Quarterly reports carry less narrative than annual reports, and the count shows it. In 2019, 564 of 17,969 10-Qs filed contained “artificial intelligence”, 3.1%.³⁴ The share passed 10% in 2023-Q4, reached 22.2% for 2025 as a whole and 32.4% in 2026-Q3, when 10-Qs for the quarter to September came in.³⁴ For 2026 to 2026-11-10, 4,796 of 14,638 10-Qs name the phrase.³⁴

The 10-Q line rises more steadily than the 10-K line because 10-Qs are spread across three quarters of the year and 10-Ks are bunched in the first.⁴ “Generative AI” stays low in 10-Qs: 6.1% of those filed in 2026-Q3.³⁴

By SIC division

EDGAR lists a Standard Industrial Classification code for most filers, and full-text search returns it with every hit.¹ Grouped into the SIC manual’s divisions, the 3,271 10-Ks filed in 2025 that name artificial intelligence come first from manufacturing (1,150), then finance, insurance and real estate (734) and services (720).¹ Transportation, communications and utilities follow with 225, retail with 132 and mining with 91.¹

Inside the divisions, single codes stand out. Pharmaceutical preparations (SIC 2834) account for 248 of the manufacturing filings and biological products (2836) for 93; prepackaged software (7372) accounts for 196 of the services filings; real estate investment trusts (6798) for 133 and state commercial banks (6022) for 122 of the finance filings, by the SEC’s code list.¹⁵ A count by division is not a share: the search caps at 10,000 hits, which rules out a whole-population query for the denominator in a single quarter, so this page reports where the filings come from, not how likely a company in each division is to use the phrase.

Where the language sits in the 10-K

A phrase count cannot say which Item a mention sits in. Item 1A risk factors are where most of the 10-K passages this site logs come from: of the 41 10-K and 10-K/A entries in the filings log, 22 quote Item 1A, 10 Item 1, 3 Item 1C and 2 Item 7, and 4 give no Item. A typical Item 1A entry puts AI next to the finance function in one sentence.⁶

The controls section is the rarest home for the language. Of the 66 entries in the log, 2 carry the CONTROLS label, both quoting Item 4 of a 10-Q, against 32 RISK FACTOR entries. The two are described in this site’s SOX testing explainer.

The queries, for anyone who wants to rerun them

Each line is one EDGAR full-text search, run once per calendar quarter from 2019-01-01 to 2026-11-10, every page read:

• “artificial intelligence”, form 10-K, and the same for form 10-Q
• “generative AI”, form 10-K, and the same for form 10-Q
• “artificial intelligence”, form 10-K, filed in 2025, with the filer’s SIC code kept

The denominators are the 32 quarterly form indexes from 2019 QTR1 to 2026 QTR4, counted for form types 10-K and 10-Q.⁴ The quarter-by-quarter counts are in the tables under each chart (“The numbers”), and the script that reruns everything is scripts/cfonews-ai/articles/research.mjs in the site’s repository. A full rerun is about 600 search requests at four a second, plus the 32 index files.

What the counts cannot tell

Three limits apply to every number above. First, a phrase hit is not a disclosure about AI in the business. Boilerplate counts: a forward-looking-statements legend listing “artificial intelligence” among risks, a cybersecurity paragraph about AI-enabled phishing, an exhibit. Second, the counts are of filings in the stated windows. They are not extrapolated to companies that do not file with the SEC, and a 10-K is counted in the quarter it was filed, not the fiscal year it covers. Third, the denominator includes every 10-K filer: operating companies, shell companies, trusts and funds that file on Form 10-K.

What the counts do show is timing. The jump from 18.0% of 10-Ks in 2023 to 36.4% in 2024 happened in one annual cycle, and the term “generative AI” went from two filings to 870 in three years.¹²

For how to read a risk factor once you have found one, see how to read an AI risk factor. How the filings log itself is assembled is on the methodology page, and the controls side of the same filings is in the controls and audit topic.