FIRC0002 Stage E (Agent 10): §1 introduction and §2 literature rewritten — contribution narrowed and 'first' claim withdrawn (LITE-01/02/08), positioning table vs Eisfeldt/Basu/Borri, AI-washing and cybersecurity strands (LITE-03/05/06), Hassan/Davis/Kim/Cao/Hong–Lim–Stein/Bloom (EDIT-14, LITE-07), §2.1 negation claim withdrawn (AI_C-07), hypotheses H3a/b, H4, H5a/b (THEO-11, ADVE-07), U8 sentences removed (ADVE-14) · 9740288 · wp17_uqo

Commit

FIRC0002 Stage E (Agent 10): §1 introduction and §2 literature rewritten — contribution narrowed and 'first' claim withdrawn (LITE-01/02/08), positioning table vs Eisfeldt/Basu/Borri, AI-washing and cybersecurity strands (LITE-03/05/06), Hassan/Davis/Kim/Cao/Hong–Lim–Stein/Bloom (EDIT-14, LITE-07), §2.1 negation claim withdrawn (AI_C-07), hypotheses H3a/b, H4, H5a/b (THEO-11, ADVE-07), U8 sentences removed (ADVE-14)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Simon-Pierre Boucher committed 4 days ago · Nov 5, 2026 parent 430ae85

2 changed files +281 −179 Browse files

modified paper/sections/introduction.tex +108 −82

`@@ -1,92 +1,118 @@`
1	% ============================================================================	1	% ============================================================================
		2	+% Owner: Stage E, Agent 10 (writing). Numbers: tables/macros.tex, sections/macros_stageE.tex.
2	\section{Introduction}	3	\section{Introduction}
3	\label{sec:introduction}	4	\label{sec:introduction}
4	% ============================================================================	5	% ============================================================================
5		6
6	−The arrival of generative artificial intelligence has been the defining corporate-narrative shift	7	+Since the public release of ChatGPT on 30~November 2022, references to large language models,
7	−of the 2020s. In the three years following the public launch of ChatGPT in late 2022, ``AI'' became	8	+generative tools, intelligent automation and the compute infrastructure behind them have spread
8	−the most frequently invoked theme on earnings calls, and the language of corporate disclosure	9	+through corporate disclosure. This paper asks whether the \emph{language} of artificial
9	−filled with references to large language models, generative tools, intelligent automation and the	10	+intelligence (AI) in firms' mandatory SEC filings is associated with their subsequent stock returns,
10	−compute infrastructure that powers them. This paper asks whether the \emph{language} of AI in	11	+when that association appears, and for which firms. It is the second version of a study whose
11	−firms' mandatory disclosures carried information about their subsequent stock returns, when that	12	+first version \citep{kammoun2026v11} answered the question for the current constituents of the
12	−information was priced, and for which firms. It is the second version of a study whose first	13	+S\&P~500 and concluded that AI language was barely priced for a decade and strongly priced in one
13	−version answered the question for the 500 current constituents of the S\&P~500 and concluded that	14	+post-launch cohort, mostly through the technology sector. The present version enlarges the
14	−AI language was barely priced for a decade and strongly priced in one post-launch cohort, mostly	15	+laboratory in three directions: to some 2,500 firms spanning the capitalization spectrum of the
15	−through the technology sector. The present version asks whether that conclusion survives when the	16	+Russell~3000, to the full text of annual and quarterly reports rather than current reports alone,
16	−laboratory is enlarged in three directions at once: to some 2,500 firms spanning the capitalization	17	+and to a second measure of AI engagement that reads what firms say they \emph{do} rather than how
17	−spectrum of the Russell~3000, to the full text of annual and quarterly reports rather than current	18	+they say it. It also incorporates the results of a peer review of the Russell~3000 version, which
18	−reports alone, and to a second measure of AI engagement that reads what firms say they \emph{do}	19	+led us to rebuild the market-capitalization variable, to clean the return series against a second
19	−rather than how they say it.	20	+price source, to replace independence-based with dependence-robust inference wherever firms share
		21	+a date, and to state every finding at the strength the corrected evidence supports.
20		22
21	−Answering the question requires three ingredients. The first is a large, machine-readable corpus of	23	+Three ingredients are needed. The first is a machine-readable corpus of disclosure: the UQO EDGAR
22	−corporate disclosure: the UQO EDGAR Research Database, extended in October 2026 to an issuer	24	+Research Database, extended in October 2026 to an issuer universe that merges the S\&P~500 snapshot
23	−universe that merges the S\&P~500 snapshot with the holdings of the iShares Russell~3000 ETF, holds	25	+with the holdings of the iShares Russell~3000 ETF, holds \mnFilingsCorpus{} Form~10-K, 10-Q and 8-K
24	−\mnFilingsCorpus{} Form~10-K, 10-Q and 8-K filings of \mnIssuersCorpus{} issuers filed since January	26	+filings of \mnIssuersCorpus{} issuers filed since January 2010, parsed into sections and
25	−2010, all of them parsed into sections and 512-word chunks. The second is a measurement technology	27	+\mChunkWordsMedian-word chunks. The second is a measurement technology that targets the meaning of
26	−that captures the meaning of text rather than the frequency of keywords: every chunk is embedded	28	+text rather than the frequency of keywords: every chunk is embedded with the
27	−with the \texttt{all-MiniLM-L6-v2} sentence transformer \citep{reimers2019sbert} and scored by	29	+\texttt{all-MiniLM-L6-v2} sentence transformer \citep{reimers2019sbert} and scored by cosine
28	−cosine similarity to written anchor passages that define three facets of AI engagement---internal	30	+similarity to written anchor passages that define three facets of AI engagement---internal
29	−adoption, infrastructure exposure, and product-disruption risk---and five placebo concepts. The	31	+adoption, infrastructure exposure and product-disruption risk---and five placebo concepts. The
30	−third is a look-ahead-free mapping from disclosure to tradeable returns: the signal of filing year	32	+third is a mapping from disclosure to tradeable returns whose timing is free of look-ahead: the
31	−$Y$ forms portfolios at the end of June of $Y{+}1$, and the forward twelve-month return runs from	33	+signal of filing year $Y$ forms portfolios at the end of June of $Y{+}1$ and the forward
32	−July of $Y{+}1$ to June of $Y{+}2$. The resulting estimation sample has \mnObs{} firm-years for	34	+twelve-month return runs from July of $Y{+}1$ to June of $Y{+}2$. The estimation sample has
33	−\mnFirms{} firms over formation years 2010--2024, of which \mnSPfy{} firm-years belong to S\&P~500	35	+\mnObs{} firm-years for \mnFirms{} firms over formation years 2010--2024, of which \mnSPfy{}
34	−members, so that every result of the earlier study can be reproduced inside the new panel and	36	+belong to S\&P~500 members, so that every result of the earlier study can be reproduced inside the
35	−placed next to its broad-universe counterpart.	37	+new panel and placed next to its broad-universe counterpart. The universe itself is a
		38	+current-membership snapshot: the timing is look-ahead-free, the selection is not, and every level
		39	+reported below is conditional on the firm surviving to October 2026 (\cref{subsec:universe,subsec:survivorship}).
36		40
37	−Three findings emerge, and each revises the earlier conclusion. First, in the broad universe AI	41	+Three findings emerge, each stated with the uncertainty the data allow. First, in the broad
38	−language carries a modest but persistent cross-sectional premium over the entire sample rather than	42	+universe AI language carries a modest cross-sectional premium over the whole sample, but its
39	−a premium that appears only after 2022. The equal-weighted top-minus-bottom quintile spread on the	43	+strength depends on the estimator. On returns winsorized within formation year, the equal-weighted
40	−AI composite is \mspreadFull{} percentage points per year ($t=\mspreadFullT$) over fifteen	44	+top-minus-bottom quintile spread on the AI composite is \mspreadFullW{} percentage points per year
41	−formation cohorts, the Fama--MacBeth slope is \mrkFMcompos{} points per standard deviation	45	+over fifteen cohorts (Newey--West $t=\mspreadFullWT$, plain $t=\mspreadFullWTplain$, 95\%
42	−($t=\mrkFMcomposT$), and a within-sector slope of \mfmFullSector{} points ($t=\mfmFullSectorT$)	46	+confidence interval \mspreadFullWCI), the median spread within cohort is \mspreadFullMed{} points,
43	−survives GICS-sector dummies. Among the \mnSmallFirms{} firms outside the largest thousand the	47	+and the raw spread of \mspreadFull{} points owes \mtopOnePctContrib{} points to the top one percent
44	−spread is \msmallFull{} points ($t=\msmallFullT$) and it is as large before the launch	48	+of returns. The Fama--MacBeth slope is more stable: \mfmComposW{} points per standard deviation
45	−(\msmallPre{}) as after (\msmallPost{}). Second, the post-ChatGPT increase documented for the	49	+(Newey--West $t=\mfmComposWT$, plain $t=\mfmComposWTplain$, interval \mfmComposWCI), and
46	−S\&P~500 replicates within the new panel---the spread among index members rises from	50	+\mfmFullSectorW{} points with GICS-sector dummies ($t=\mfmFullSectorWT$). The premium is therefore
47	−\mspSpreadPre{} to \mspSpreadPost{} points, with the 2024 cohort again dominant---but it does not	51	+robust in regression form and fragile in sort form; within sectors it survives in the regressions
48	−generalize to the market. In the Russell~3000 the post-launch spread is \mspreadPost{} points	52	+and not in a sector-neutral quintile sort (\msectNeuFullW{} points, $t=\msectNeuFullWT$). Second,
49	−against \mspreadPre{} before ($t$ of the post mean \mspreadPostT), the 2022 cohort earned a	53	+the post-ChatGPT increase documented for the S\&P~500 replicates within the new panel---the
50	−negative spread, dropping the 2024 cohort leaves a post-period mean below the pre-period one, a	54	+winsorized spread among index members rises from \mspSpreadPreW{} to \mspSpreadPostW{} points,
51	−placebo over ten candidate break years ranks 2022 \mplaceboRank{}th of \mplaceboN{} (the largest	55	+with the 2024 cohort again dominant---but it does not generalize to the market. In the
52	−break candidate is \mplaceboBest), and the monthly long--short portfolio earns a post-launch	56	+Russell~3000 the post-launch spread is \mspreadPostW{} points against \mspreadPreW{} before, on
53	−six-factor alpha of \mffsixPost\% per year. The increase is a large-cap phenomenon: among the	57	+three cohorts whose interval is \mspreadPostWCI; the 2022 cohort is negative; the pooled
54	−largest thousand firms the spread rises from \mlargePre{} to \mlargePost{} points	58	+post-minus-pre difference is \mpooledDiffW{} points (firm-clustered $t=\mpooledDiffWT$, two-way
55	−($t=\mlargePostT$), and in the ChatGPT event study the largest thousand firms in the top tercile of	59	+$t=\mpooledDiffWTtw$); the 2022--24 block ranks no higher than a random three-cohort block
56	−pre-launch AI language earned \mevLarge{} points more one-year abnormal return than the bottom	60	+(permutation $p=\mpermPW$); and the monthly long--short earns a post-launch six-factor alpha of
57	−tercile ($t=\mevLargeT$) against \mevSmall{} ($t=\mevSmallT$) for the remaining firms. Third, the	61	+\mffsixPost\% per year. An increase is visible only among the largest thousand firms, where the
58	−two new measurement dimensions matter. The signal built from annual reports alone, which the	62	+winsorized spread rises from \mlargePreW{} to \mlargePostW{} points, and only because of the 2024
59	−earlier study could not compute, is the strongest of the three form-specific signals (full-sample	63	+cohort: without it the large-cap post-launch mean is \mlargePostExTwentyFourW{} points. The formal
60	−spread \mtenkPost{} points after the launch, Fama--MacBeth slope \mhrTenK{} per standard deviation	64	+test of the large-versus-small difference---a pooled regression with a high-AI $\times$ post
61	−in a horse race against the quarterly- and current-report signals), whereas the count of AI,	65	+$\times$ large interaction---gives \mcontrastLarge{} points with a firm-clustered $t$ of
62	−data-center, cloud and automation \emph{projects} extracted from the same filings by the SPID	66	+\mcontrastLargeT{}, a two-way $t$ of \mcontrastLargeTtw{} and a wild-cluster bootstrap $p$ of
63	−database carries no premium at all (\mspidPre{} and \mspidPost{} points before and after the	67	+\mcontrastLargePwild{}, against a minimum detectable effect of \mcontrastLargeMDE{} points: the
64	−launch).	68	+gradient is visible in the point estimates and not established by the test. The ChatGPT event
		69	+study shows the same pattern and the same fragility: under the single-index market model the
		70	+largest thousand firms in the top tercile of pre-launch AI language earned \mevLarge{} points more
		71	+one-year abnormal return than the bottom tercile (dependence-robust $t=\mevLargeCDA$) against
		72	+\mevSmall{} ($t=\mevSmallCDA$) for the remaining firms, but under beta-only, market-adjusted,
		73	+size-and-momentum-matched and six-factor calendar-time benchmarks the large-cap statistic falls
		74	+to \mevLargeBetaCDA, \mevLargeMadjCDA, \mevLargeSizeMomCDA{} and \mevLargeFFsixT. Third, the two
		75	+new measurement dimensions matter, with one caveat each. The signal built from annual reports has
		76	+the largest slope in a horse race against the quarterly- and current-report signals (\mhrTenKW{}
		77	+points per standard deviation, $t=\mhrTenKWT$), but the ranking of the three forms depends on the
		78	+specification: a Wald test rejects equal slopes on winsorized returns ($p=\mformWaldP$) and not on
		79	+raw returns ($p=\mformWaldPraw$). And the count of AI,
		80	+data-center, cloud and automation \emph{projects} that a large-language-model reader extracts from
		81	+the same filings shows no detectable premium: firms with at least one AI-initiative project earn
		82	+\mspidAnyFirmFull{} points more than other firms ($t=\mspidAnyFirmFullT$, interval
		83	+\mspidAnyFirmFullCI), with a minimum detectable effect of \mspidAnyFirmFullMDE{} points on a
		84	+measure whose precision has not yet been validated against human labels.
65		85
66	−Taken together, the evidence is consistent with a persistent and partly within-industry premium on	86	+Taken together, the evidence is consistent with a modest, regression-robust association between
67	−AI exposure in the broad market, present throughout the 2010s among smaller firms, together with a	87	+AI language and subsequent returns in the broad market, conditional on survival, and with a
68	−re-rating of the largest, most-followed AI-exposed firms after 2022 that the S\&P~500 sample	88	+post-2022 increase that is confined to the largest firms and to one cohort and whose event-study
69	−isolated and that an attention-based account of the post-launch period would not predict. The	89	+expression depends on the benchmark; it is not consistent with a market-wide regime change. The
70	sector channel that dominated the earlier study is still visible---the long--short loads	90	sector channel that dominated the earlier study is still visible---the long--short loads
71	−\mtechBetaPost{} on a technology-sector factor after the launch, sector-by-year fixed effects	91	+\mtechBetaPost{} on a technology-sector factor after the launch, and sector-by-year fixed effects
72	−absorb the post-launch cross-sectional slope entirely, and cybersecurity language is rewarded as	92	+reduce the post-launch cross-sectional slope to \msyPost{} points ($t=\msyPostT$)---but over the
73	−much as AI language---but in the broad universe it coexists with a within-sector premium that the	93	+full sample it coexists with a within-sector slope that the index sample was too small to detect.
74	−index sample was too small and too homogeneous to detect.	94	+We also report what the measure is and is not: the encoder reads about \mTruncWindowWords{} of
		95	+the \mChunkWordsMedian{} words of a typical chunk, a keyword count correlates only \mKwRho{} with
		96	+the composite, and two electric utilities sit in the top dozen firms, so the ``AI language''
		97	+construct is best read as proximity to a technology-and-infrastructure vocabulary rather than as
		98	+validated AI activity (\cref{subsec:construct}).
75		99
76	−\paragraph{Contribution.} The paper makes three contributions. First, substantively, it provides	100	+\paragraph{Contribution.} Relative to the text-based AI-exposure measures of
77	−the first look-ahead-free evidence on the pricing of AI disclosure language across the full	101	+\citet{eisfeldt2026generative}, \citet{basu2026aiexposure} and \citet{borri2026aifactor}, which
78	−capitalization spectrum of the U.S.\ market, and shows that the headline post-ChatGPT result of the	102	+rely on workforce task content, earnings-call language or realized model usage, we measure
79	−S\&P~500 literature is a large-cap result. Second, methodologically, it demonstrates that	103	+exposure from the firm's own mandatory filings across all three forms for the Russell~3000, trace
80	−embedding-based disclosure measurement at the scale of 2,500 firms and 527,000 filings is feasible,	104	+its association with returns over fifteen look-ahead-free cohorts on a current-membership universe
81	−documents how the signal depends on the form of the filing it is computed from, and shows that a	105	+rather than over an event window, and show that the post-launch premium documented in index
82	−language measure and a project-count measure extracted from the same text can have opposite	106	+samples is confined to the largest thousand firms; and relative to the AI-washing literature
83	−pricing implications. Third, it provides a template for disciplining regime claims built on short	107	+\citep{erdogan2026aiwashing,wang2026aiwashing} we show that a language-model count of disclosed AI
84	−post-event samples---cohort-by-cohort and leave-one-out estimates, pooled firm-clustered tests,	108	+projects carries no detectable premium while the language does. Methodologically, the paper
85	−placebo break dates, sector-factor spanning, winsorization and dependence-robust event-study	109	+documents how an embedding signal depends on the filing form it is computed from and on the
86	−inference---and applies it symmetrically to the index sample and to the broad universe. The	110	+encoder's input window, and it applies to its own regime claim the discipline that short
87	−remainder of the paper is organized as follows. Section~\ref{sec:literature} situates the study and	111	+post-event samples require: cohort-by-cohort and leave-one-out estimates, two-way clustering and
88	−states the hypotheses. Section~\ref{sec:data} describes the corpus, the universe, the semantic	112	+wild-cluster bootstraps, permutation and sup-Wald break tests, multiple-testing adjustments across
89	−measures and the timing convention. Section~\ref{sec:methodology} lays out the designs.	113	+families, placebo launches and dependence-robust event-study inference under five benchmarks.
90	−Section~\ref{sec:results} presents the main results for both universes,	114	+\Cref{sec:literature} positions the study and states the hypotheses; \cref{sec:data} describes
91	−Section~\ref{sec:robustness} the robustness and identification tests, Section~\ref{sec:discussion}	115	+the corpus, the universe, the measures and the timing; \cref{sec:methodology} the designs and the
92	−the interpretation and limitations, and Section~\ref{sec:conclusion} concludes.	116	+inference; \cref{sec:results} the main results for both universes; \cref{sec:robustness} the
		117	+robustness, construct-validity and survivorship diagnostics; \cref{sec:discussion} the
		118	+interpretation and limitations; \cref{sec:conclusion} concludes.

modified paper/sections/literature.tex +173 −97

`@@ -1,41 +1,57 @@`
1	% ============================================================================	1	% ============================================================================
		2	+% Owner: Stage E, Agent 10 (writing). Numbers: tables/macros.tex, sections/macros_stageE.tex.
2	\section{Related Literature and Hypotheses}	3	\section{Related Literature and Hypotheses}
3	\label{sec:literature}	4	\label{sec:literature}
4	% ============================================================================	5	% ============================================================================
5		6
6	−Our study sits at the intersection of five strands of research: textual analysis in finance, the	7	+The study draws on six strands of research: textual analysis in finance, the pricing of corporate
7	−pricing of corporate disclosure, technology and innovation in asset pricing, investor attention and	8	+disclosure, text-based measures of technology and AI exposure, investor attention, the AI-washing
8	−narrative economics, and the nascent literature on the economics of artificial intelligence. Each	9	+and cybersecurity-disclosure literatures that bear on what the measure captures, and the
9	−contributes a building block---a measurement tool, an object of study, a candidate risk factor, a	10	+methodological literature on short samples, multiple testing and event-study inference. We take
10	−behavioural mechanism, and a comparison case---and a sixth, methodological strand disciplines the	11	+the strands in turn, place the paper against its three closest competitors, and close with the
11	−inference. We take the strands in turn and close with the hypotheses they jointly motivate.	12	+hypotheses.
12		13
13	\subsection{Textual analysis and natural language processing in finance}	14	\subsection{Textual analysis and natural language processing in finance}
		15	+\label{subsec:lit_text}
14	A now-large literature extracts economic signals from financial text. \citet{tetlock2007} showed	16	A now-large literature extracts economic signals from financial text. \citet{tetlock2007} showed
15	that the tone of a \textit{Wall Street Journal} column forecasts aggregate returns, and	17	that the tone of a \textit{Wall Street Journal} column forecasts aggregate returns, and
16	\citet{tetlock2008} extended this to firm-level news. \citet{lm2011} demonstrated that	18	\citet{tetlock2008} extended this to firm-level news. \citet{lm2011} demonstrated that
17	finance-specific sentiment dictionaries materially outperform general-purpose word lists in 10-K	19	finance-specific sentiment dictionaries materially outperform general-purpose word lists in 10-K
18	−analysis, launching a wave of dictionary-based studies; \citet{lm2016} and \citet{lm2020} survey	20	+analysis; \citet{lm2016} and \citet{lm2020} survey the field and its pitfalls, including the
19	−the field and its pitfalls, including the sensitivity of results to document selection and	21	+sensitivity of results to document selection and pre-processing. A second generation moved from
20	−pre-processing. A second generation moved from counting words to modelling them: \citet{gentzkow2019}	22	+counting words to modelling them: \citet{gentzkow2019} review text-as-data methods in economics,
21	−review text-as-data methods in economics, \citet{kekellyxiu2019} develop a supervised topic model	23	+\citet{kekellyxiu2019} develop a supervised topic model that predicts returns from news,
22	−that predicts returns from news, \citet{kelly2019textregression} propose a sparse text-regression	24	+\citet{kelly2019textregression} propose a sparse text-regression framework, and
23	−framework, and \citet{bybee2024narrative} recover a ``state of the economy'' from business news.	25	+\citet{bybee2024narrative} recover a ``state of the economy'' from business news. A third
24	−A third generation uses distributed representations of meaning. \citet{hanleyhoberg2019} apply word	26	+generation uses distributed representations of meaning. \citet{hanleyhoberg2019} apply word
25	embeddings to bank 10-Ks to identify emerging risks before they are named in dictionaries;	27	embeddings to bank 10-Ks to identify emerging risks before they are named in dictionaries;
26	\citet{limai2021} measure corporate culture from earnings calls with word embeddings;	28	\citet{limai2021} measure corporate culture from earnings calls with word embeddings;
27	\citet{huang2023finbert} show that a domain-adapted transformer outperforms dictionaries and	29	\citet{huang2023finbert} show that a domain-adapted transformer outperforms dictionaries and
28	bag-of-words classifiers on financial text; and \citet{cong2022textfactor} build interpretable	30	bag-of-words classifiers on financial text; and \citet{cong2022textfactor} build interpretable
29	−``textual factors'' from embeddings. Our paper belongs to this embedding-based generation. The	31	+``textual factors'' from embeddings. The firm-level exposure measures of \citet{hassan2019firmlevel}
30	−limitations of keyword counting are acute for a theme like AI: a count cannot distinguish ``we do	32	+(political risk from earnings-call text), \citet{davis2020exposures} (COVID-19 exposures),
31	−not currently use artificial intelligence'' from ``we are an artificial-intelligence leader,'' it	33	+\citet{bloom2021diffusion} (technology diffusion from patents, calls and job postings) and
32	−misses the paraphrase of a firm that describes machine learning without writing ``AI,'' and it	34	+\citet{kim2023llmrisks} (risk assessments produced by a generative model) are the direct
33	−treats boilerplate and substance identically. We instead measure the semantic similarity of each	35	+methodological lineage of a construct-specific score computed from a firm's own text.
34	−passage to anchor concepts with a sentence transformer \citep{reimers2019sbert}, a design close in	36	+
35	−spirit to the similarity-to-reference-text constructions of \citet{hanleyhoberg2019}. The	37	+Our paper belongs to the embedding-based generation. Its measure is the semantic similarity of
36	−literature also warns, however, that text-based measures inherit the composition of the corpus they	38	+each passage to anchor paragraphs, computed with a sentence transformer
37	−are computed on \citep{lm2016,lm2020}; Section~\ref{sec:data} therefore documents that composition	39	+\citep{reimers2019sbert}---a design close in spirit to the similarity-to-reference-text
38	−in detail.	40	+constructions of \citet{hanleyhoberg2019} and \citet{celenymarechal2024cyber}. Embeddings have
		41	+two advantages over a keyword count: a passage that describes machine learning without writing
		42	+``AI'' is scored, and a count of the acronym does not distinguish a risk-factor boilerplate from
		43	+an operating description. They also have limits that \cref{subsec:construct} measures rather
		44	+than assumes. The first version of this paper claimed that the embedding separates ``we do not
		45	+currently use artificial intelligence'' from ``we are an artificial-intelligence leader''; a
		46	+direct probe shows that it does so only in part---negated sentences score lower than affirmative
		47	+ones (\mProbeNegated{} against \mProbeAffirm{} on the adoption anchor) but still above the
		48	+\mProbeNegatedPct{}th percentile of corpus chunks---so the claim is withdrawn. Sentence encoders
		49	+of the MiniLM class are known to be weakly sensitive to negation, and the measure should be read
		50	+as topical proximity to AI vocabulary rather than as a stance. The literature also warns that
		51	+text-based measures inherit the composition of the corpus they are computed on
		52	+\citep{lm2016,lm2020} and that firms adapt their language to the machines that read it
		53	+\citep{cao2023machinelistening}; \cref{sec:data,subsec:construct} therefore document the corpus
		54	+composition, the encoder's input window and the keyword baseline in detail.
39		55
40	\subsection{The pricing of corporate disclosure}	56	\subsection{The pricing of corporate disclosure}
41	A parallel literature asks whether the \emph{content} and \emph{changes} of mandatory disclosure	57	A parallel literature asks whether the \emph{content} and \emph{changes} of mandatory disclosure
`@@ -43,35 +59,78 @@ predict returns. \citet{cohen2020lazy} show that firms that change the language`
43	reports relative to the prior year experience predictable future returns and operating outcomes.	59	reports relative to the prior year experience predictable future returns and operating outcomes.
44	\citet{hoberg2010product} and \citet{hobergphillips2016} use 10-K product descriptions to build	60	\citet{hoberg2010product} and \citet{hobergphillips2016} use 10-K product descriptions to build
45	text-based industry classifications that carry asset-pricing content. \citet{lopezlira2023risk}	61	text-based industry classifications that carry asset-pricing content. \citet{lopezlira2023risk}
46	−extracts priced risk exposures from risk-factor sections. The general lesson is that the	62	+extracts priced risk exposures from risk-factor sections. The lesson is that the qualitative
47	−qualitative portions of filings contain value-relevant information that markets process	63	+portions of filings contain value-relevant information that markets process incompletely and with
48	−incompletely and with delay. We contribute a concept-level measure---AI engagement---computed from the annual, quarterly	64	+delay. We contribute a concept-level measure---AI engagement---computed from the annual,
49	−and current reports of a broad universe of U.S.\ listed firms, and we characterize how its relation to	65	+quarterly and current reports of a broad universe of U.S.\ listed firms, and we characterize how
50	−returns varies over time and across firm size, dimensions the disclosure-pricing literature has largely left implicit.	66	+its association with returns varies over time and across firm size, two dimensions the
		67	+disclosure-pricing literature has largely left implicit.
		68	+
		69	+\subsection{Text-based AI exposure and the cross-section of returns: the closest papers}
		70	+\label{subsec:lit_ai}
		71	+Asset pricing has long sought to connect technological change to risk and return
		72	+\citep{hirshleifer2013innovative,kogan2017technological,koganpapanikolaou2014}, and a recent
		73	+literature measures AI exposure directly. \citet{babina2024aiinvestment} measure firm-level AI
		74	+investment from job postings and r\'esum\'es and document faster growth among AI-adopting firms;
		75	+\citet{acemoglu2022aijobs}, \citet{webb2020} and \citet{hampole2025ai} measure the exposure of
		76	+occupations, establishments and firms' workforces to AI; \citet{acemoglu2020wrong} study the
		77	+labour-market effects of industrial robots. Three papers of 2026 are the closest to ours and were
		78	+absent from the first version. \citet{eisfeldt2026generative} construct a firm-level measure of
		79	+\emph{generative}-AI exposure from the task content of occupations, show that a long--short
		80	+portfolio on that measure earned large returns in the two weeks after ChatGPT's release, and
		81	+separate this workforce exposure from firms' \emph{product} exposure to AI. \citet{basu2026aiexposure}
		82	+build text-based AI-exposure measures from the earnings-call transcripts of some 13,800 firms in
		83	+44 countries over 2010--2025, find that a one-standard-deviation change in exposure is associated
		84	+with higher annual returns concentrated after the ChatGPT launch, and report that a U.S.\
		85	+long--short portfolio of high-exposure firms earns no abnormal return before the launch, once its
		86	+large-cap growth tilts are accounted for, and a large one afterward. \citet{borri2026aifactor}
		87	+measure exposure from realized consumption of language-model tokens and find a positive premium
		88	+in developed markets. Three further papers bear on specific constructs: \citet{song2026productexposure}
		89	+builds a product-exposure measure from 10-K business descriptions, which is the object our
		90	+disruption-risk anchor targets; \citet{wulam2026genai} show that short-window reactions to AI
		91	+discussions on earnings calls depend on whether the discussion conveys new or concrete
		92	+information, a result that parallels our language-versus-projects contrast; and
		93	+\citet{jha2024chatgpt} and \citet{lopezliratang2026} use a language model itself as the reader of
		94	+corporate or news text. \citet{yilmazkuday2026chatgpt} examines the aggregate return response to
		95	+ChatGPT.
		96	+
		97	+\Cref{tab:positioning} places the paper against the three closest papers. Our measure is built
		98	+from the firm's own mandatory filings across all three forms, which none of them uses; our return
		99	+profile covers fifteen look-ahead-free formation cohorts on a current-membership universe rather
		100	+than an event window; and our second measure---a language-model count of disclosed AI
		101	+projects---has no counterpart. Where \citet{basu2026aiexposure} find the association concentrated
		102	+after the launch, we find a modest association throughout the sample in regression form and a
		103	+post-launch increase confined to the largest firms; the two readings are not incompatible, because
		104	+their sample is dominated by the large firms in which we also find the increase, but the
		105	+difference in universe, text source and horizon should be kept in mind. Where
		106	+\citet{eisfeldt2026generative} and \citet{song2026productexposure} distinguish labour from product
		107	+exposure, our three anchors map imperfectly onto that distinction: the adoption anchor describes
		108	+the firm's own use of AI (closest to an operating, partly labour-saving exposure), the
		109	+infrastructure anchor describes compute and data-center investment (a supplier or user position
		110	+in the AI build-out, which neither measure targets), and the disruption-risk anchor describes the
		111	+threat of AI to the firm's products (the product exposure of Song). The sign and timing of our
		112	+disruption-risk spread are compared with their results in \cref{sec:discussion}.
51		113
52	−\subsection{Technology, innovation, and the cross-section of returns}	114	+\begin{table}[htbp]\centering\footnotesize
53	−Asset pricing has long sought to connect technological change to risk and return.	115	+\caption{\textbf{Positioning relative to the closest papers.} Qualitative summary of each paper's
54	−\citet{hirshleifer2013innovative} show that innovative efficiency predicts returns;	116	+question, measure, sample and headline finding, taken from the published abstracts; no number of
55	−\citet{kogan2017technological} build market-based measures of innovation from patents and link them	117	+this paper is compared here.}\label{tab:positioning}
56	−to growth and reallocation; \citet{koganpapanikolaou2014} show that investment-specific technology	118	+\begin{tabular}{@{}p{3.1cm}p{3.6cm}p{3.1cm}p{5.6cm}@{}}
57	−shocks shape the cross-section. A more recent literature studies AI exposure directly.	119	+\toprule
58	−\citet{babina2024aiinvestment} measure firm-level AI investment from job postings and r\'esum\'es	120	+Paper & Exposure measure & Sample & Headline finding and relation to this paper \\
59	−and document faster growth among AI-adopting firms; \citet{acemoglu2022aijobs} and \citet{webb2020}	121	+\midrule
60	−measure the exposure of occupations and establishments to AI; \citet{acemoglu2020wrong} study the	122	+\citet{eisfeldt2026generative} & Occupational task content of the workforce (generative-AI exposure); product exposure separated & U.S.\ listed firms, event window around 30 Nov 2022 & Long--short on workforce exposure earns large returns in the two weeks after the launch; labour and product exposures distinct. We use filing language, fifteen annual cohorts and a size split; our disruption-risk anchor is a product-exposure construct. \\
61	−labour-market effects of industrial robots. Closest to us, \citet{eisfeldt2023generative} construct	123	+\addlinespace
62	−a firm-level measure of \emph{generative}-AI exposure from the task content of occupations and show	124	+\citet{basu2026aiexposure} & Hybrid NLP on earnings-call transcripts; topic-specific AI measures & $\approx$13,800 firms, 44 countries, 2010--2025 & Return association concentrated after ChatGPT; U.S.\ long--short has no pre-launch alpha once large-cap growth tilts are controlled. Same question and event; we use mandatory filings, a U.S.\ Russell~3000 universe and a language-versus-projects contrast, and find the post-launch increase only among the largest firms. \\
63	−that a long--short portfolio on that measure earned large returns in the weeks after ChatGPT's	125	+\addlinespace
64	−release; \citet{jha2024chatgpt} use ChatGPT itself to extract expected investment and employment	126	+\citet{borri2026aifactor} & Realized consumption of language-model tokens & U.S.\ and international equities & Positive premium on usage-based exposure, significant in developed markets. A third exposure construct; usage rather than disclosure. \\
65	−policies from earnings calls; and \citet{lopezliratang2026} show that a large language model can	127	+\addlinespace
66	−rank news headlines in a way that predicts next-day returns. Our contribution relative to this work	128	+This paper & Sentence-embedding similarity of 10-K, 10-Q and 8-K text to anchor passages; LLM count of disclosed AI projects & \mnFirms{} firms (current S\&P~500 $\cup$ IWV holdings), formation cohorts 2010--2024 & Modest regression-robust, sort-fragile association over the full sample, conditional on survival; post-launch increase confined to the largest thousand firms and the 2024 cohort; no detectable premium on disclosed projects. \\
67	−is to measure AI exposure from the firm's own disclosure language, available for some 2,500 firms	129	+\bottomrule
68	−at annual frequency, and to characterize the full time profile of its pricing across fifteen	130	+\end{tabular}
69	−cohorts rather than an event window. Where \citet{eisfeldt2023generative} document a sharp	131	+\end{table}
70	−repricing in the weeks around the launch, we ask whether a persistent, text-based exposure measure
71	−carried a premium before the launch and whether the premium after it is distinguishable from a
72	−technology-sector effect.
73		132
74	−\subsection{Investor attention, salience, and narrative economics}	133	+\subsection{Investor attention, salience, and the size gradient}
75	Why might value-relevant disclosure be priced only after a catalyst? The behavioural literature	134	Why might value-relevant disclosure be priced only after a catalyst? The behavioural literature
76	offers a mechanism: limited investor attention. \citet{barberodean2008} show that individual	135	offers a mechanism: limited investor attention. \citet{barberodean2008} show that individual
77	investors are drawn to attention-grabbing stocks; \citet{da2011} construct an attention proxy from	136	investors are drawn to attention-grabbing stocks; \citet{da2011} construct an attention proxy from
`@@ -79,60 +138,77 @@ Google search and link it to prices; \citet{dellavigna2009} document under-react`
79	released when attention is low; and \citet{hirshleifer2009driven} show that earnings news is	138	released when attention is low; and \citet{hirshleifer2009driven} show that earnings news is
80	underweighted when investors are distracted. \citet{shiller2017narrative} argues that contagious	139	underweighted when investors are distracted. \citet{shiller2017narrative} argues that contagious
81	narratives drive major asset-price movements, and \citet{bakerwurgler2006} show that sentiment	140	narratives drive major asset-price movements, and \citet{bakerwurgler2006} show that sentiment
82	−differentially affects hard-to-value, speculative stocks. These mechanisms predict that a salient	141	+differentially affects hard-to-value, speculative stocks. \citet{honglimstein2000} show that
83	−consumer event such as ChatGPT could draw attention to information that had been sitting in filings	142	+firm-specific information diffuses more slowly among small, lightly covered firms, which is the
84	−for years. They do not, however, distinguish a firm-level attention shock from a sector-level	143	+natural prediction for where a disclosure-based signal should be priced with the longest delay.
85	−revision of expected cash flows, and our tests are designed to separate the two.	144	+These mechanisms predict that a salient consumer event such as ChatGPT could draw attention to
		145	+information that had been sitting in filings for years, and that the effect should be larger
		146	+where attention frictions are larger. They do not distinguish a firm-level attention shock from a
		147	+sector-level revision of expected cash flows, and our designs separate the two only partially:
		148	+no attention proxy is in the panel, so the size gradient is the only attention-related test.
86		149
87	−\subsection{ESG and climate disclosure as a comparison case}	150	+\subsection{AI washing, cybersecurity disclosure, and what a language measure captures}
88	−Our placebo design connects to the literature on the pricing of ESG and climate disclosure.	151	+\label{subsec:lit_washing}
89	−\citet{pastor2021sustainable} model how sustainability preferences depress the expected returns of	152	+Two literatures bear on the interpretation of a premium on AI \emph{language} that is absent on
90	−green assets in equilibrium, and \citet{pastor2022dissecting} show that green stocks' strong	153	+AI \emph{projects}. The first names the phenomenon ``AI washing'': \citet{erdogan2026aiwashing}
91	−realized returns through 2021 reflected unexpected increases in climate concern rather than high	154	+study the disclosure response of U.S.\ firms to the SEC's warnings about unsupported AI claims,
92	−expected returns. \citet{bolton2021risk} document a carbon premium, and \citet{sautner2023} build a	155	+and \citet{wang2026aiwashing} find that AI washing earns a short-lived post-disclosure
93	−firm-level climate-exposure measure from earnings-call language that is priced in options and	156	+abnormal-return premium that later reverses. Our reading of the language premium as a premium on
94	−returns. Consistent with the post-2021 reversal in this literature, we find that high-ESG and	157	+``sounding like an AI firm'' is the AI-washing hypothesis, and it carries a testable
95	−high-climate disclosure earned negative spreads after 2022, which helps to rule out a generic	158	+implication---reversal at longer horizons---that \cref{subsec:spidres} examines. The second
96	−``more disclosure earns more'' explanation of the AI result.	159	+documents a cybersecurity premium: \citet{celenymarechal2024cyber} measure the proximity of 10-K
		160	+text to a cyber corpus with a construction like ours and find that high-cyber-risk stocks earn
		161	+higher returns, and \citet{kearney2025cyberchatgpt} study cyber risk around the ChatGPT launch.
		162	+The cybersecurity spread that our placebo design produces is therefore a documented premium, not
		163	+an unexplained artefact, and \cref{subsec:placebo,subsec:spidres} report the correlation of the
		164	+two measures and a cyber-neutral AI sort. The ESG and climate placebos connect to
		165	+\citet{pastor2021sustainable}, \citet{pastor2022dissecting}, \citet{bolton2021risk} and
		166	+\citet{sautner2023}; consistent with the post-2021 reversal in that literature, high-ESG and
		167	+high-climate language earned negative spreads after 2022.
97		168
98	−\subsection{Short samples, data snooping, and event-study inference}	169	+\subsection{Short samples, multiple testing, survivorship, and event-study inference}
99	−Two methodological literatures bear directly on how much weight our post-2022 estimates can carry.	170	+Four methodological literatures govern how much weight the estimates can carry.
100	−The first concerns the credibility of cross-sectional return patterns. \citet{harveyliuzhu2016}	171	+\citet{harveyliuzhu2016} argue that the multiplicity of tested predictors requires $t$-statistics
101	−argue that the multiplicity of tested predictors requires $t$-statistics well above conventional	172	+well above conventional thresholds; \citet{houxuezhang2020} find that most published anomalies
102	−thresholds; \citet{houxuezhang2020} find that most published anomalies fail to replicate under	173	+fail to replicate under standard procedures; and \citet{mcleanpontiff2016} document
103	−standard procedures; and \citet{mcleanpontiff2016} document post-publication decay. A spread	174	+post-publication decay. A spread estimated on three annual observations, one of which is extreme,
104	−estimated on three annual observations, one of which is extreme, must be read in that light. The	175	+must be read in that light, and \cref{subsec:mht} reports family sizes, adjusted $p$-values, a
105	−second concerns inference in event studies when all firms share a single event date, so that	176	+deflated Sharpe ratio and a specification curve. The use of a current-membership universe raises
106	−abnormal returns are cross-sectionally correlated \citep{brownwarner1985,kolari2010} and	177	+the survivorship and look-back selection analysed by \citet{browngoetzmannross1995}; the first
107	−event-induced variance may differ from estimation-window variance \citep{boehmer1991}, and in	178	+version addressed it with restrictions on listing history, which condition on entry, not on exit,
108	−long-horizon settings, where the choice of benchmark and the compounding of abnormal returns	179	+and \cref{subsec:survivorship} replaces them with point-in-time holdings, attrition rates and a
109	−matter \citep{barberlyon1997,fama1998,mitchellstafford2000,kothariwarner2007}. Our event-study	180	+bounding exercise. Event studies in which all firms share a single date have cross-sectionally
110	−inference follows the portfolio approach recommended in that literature. Finally, the use of a	181	+correlated abnormal returns \citep{brownwarner1985,kolari2010}, event-induced variance
111	−current-membership universe raises the survivorship concerns analysed by	182	+\citep{boehmer1991} and, at long horizons, a dependence on the benchmark and the compounding
112	−\citet{browngoetzmannross1995}, which we address with restrictions on listing history.	183	+convention \citep{barberlyon1997,fama1998,mitchellstafford2000,kothariwarner2007}; our inference
		184	+follows the portfolio approach of that literature and reports every statistic under several
		185	+benchmarks. Finally, inference in panels with few time clusters follows
		186	+\citet{petersen2009}, \citet{thompson2011} and \citet{cameron2011}.
113		187
114	\subsection{Hypotheses}	188	\subsection{Hypotheses}
115	\label{subsec:hypotheses}	189	\label{subsec:hypotheses}
116	−Our analysis is organized around four hypotheses.	190	+The analysis is organized around five hypotheses, each stated as a prediction about an
		191	+association and each with the clause that would falsify it.
117	\begin{description}[leftmargin=2.4em,style=nextline]	192	\begin{description}[leftmargin=2.4em,style=nextline]
118	−\item[H1 (Unconditional pricing).] If AI exposure carries a stable risk premium or persistent	193	+\item[H1 (Unconditional association).] If AI exposure carries a stable premium or persistent
119	−mispricing, then sorting on AI-disclosure intensity yields a positive and significant risk-adjusted	194	+mispricing, sorting on AI-disclosure intensity yields a positive risk-adjusted return spread over
120	−return spread over the full 2010--2024 formation sample.	195	+the full 2010--2024 formation sample, in sort and in regression form, that survives winsorization.
121	\item[H2 (Regime dependence).] If instead AI exposure is repriced only once AI becomes salient, the	196	\item[H2 (Regime dependence).] If instead AI exposure is repriced only once AI becomes salient, the
122	−return spread is small before and large after the late-2022 generative-AI catalyst, and the	197	+return spread is small before and large after the late-2022 generative-AI catalyst, the increase
123	−increase is visible in each post-launch cohort rather than in a single one.	198	+is visible in each post-launch cohort rather than in a single one, and the 2022 break date stands
		199	+out among placebo break dates.
124	\item[H3 (AI-specificity).] If the post-2022 spread reflects an AI repricing rather than a	200	\item[H3 (AI-specificity).] If the post-2022 spread reflects an AI repricing rather than a
125	−mechanical or market-wide artefact, it appears for AI-related semantic factors but not for unrelated	201	+mechanical or market-wide artefact, (a) it appears for AI-related semantic factors but not for
126	−factors such as ESG, climate, or macro risk, and it is not spanned by a technology-sector factor.	202	+unrelated factors such as ESG, climate or macro risk, and (b) it is not spanned by a
		203	+technology-sector factor.
127	\item[H4 (Cross-sectional, not within-firm).] If the market repriced \emph{which} firms are	204	\item[H4 (Cross-sectional, not within-firm).] If the market repriced \emph{which} firms are
128	−AI-exposed rather than rewarding incremental wording, the effect is concentrated in the	205	+AI-exposed rather than rewarding incremental wording, the association is concentrated in the
129	−cross-section, attenuates under firm fixed effects, and displays no pre-trend when exposure is held	206	+cross-section, attenuates under firm fixed effects, and the event-time profile of a pre-determined
130	−fixed at its pre-launch level.	207	+exposure shows no differential trend before the launch and a positive shift after it.
131	−\end{description}	208	+\item[H5 (Size and projects).] (a) If the post-launch repricing reflects limited attention rather
132	−\noindent We add a fifth hypothesis that the broad universe makes testable.	209	+than a revision of sector cash flows, it is stronger among smaller, less-followed firms than among
133	−\begin{description}[leftmargin=2.4em,style=nextline]	210	+the largest thousand. (b) If the language premium reflects concrete AI activity, it is visible in
134	−\item[H5 (Size and attention).] If the post-launch repricing reflects limited attention rather than a	211	+the firm's own disclosed AI projects as well as in its language.
135	−revision of sector cash flows, it is stronger among smaller, less-followed firms than among the largest
136	−thousand, and it is visible in the firm's own disclosed AI projects as well as in its language.
137	\end{description}	212	\end{description}
138	−\noindent Section~\ref{sec:discussion} summarizes what the Russell~3000 evidence says about each hypothesis.	213	+\noindent \Cref{subsec:verdicts} records, clause by clause, what the Russell~3000 evidence says
		214	+about each hypothesis.
139		215

spb wp17_uqo

Theme