Weasel words in an annual report are qualifiers that let a claim sound authoritative while committing the company to little that can be verified. Typical examples are modal verbs (may, could, might), vague adjectives and adverbs (approximately, generally, substantial), and distancing phrases (we believe, it appears). Hedging is not deception in itself. The analytical task is separating caution that protects accuracy from caution that replaces specificity.
TL;DR
- Hedging terms fall into recognizable classes: modal verbs, hedging adjectives, adverbs and nouns, and cognitive modifiers such as "we believe" and "in our view."
- Hedging alone does not signal deception. U.S. law encourages forward-looking statements with cautionary language, so the test is whether a qualifier sits on a genuine uncertainty or stands in for a fact.
- Research on 10-Ks links obfuscating language, low readability and ambiguous tone to higher borrowing costs, higher equity costs and, in some samples, fraudulent reporting.
- A reliable assessment compares each qualified claim with checkable records: MD&A, audited statements and notes, controls disclosures, and non-GAAP reconciliations on EDGAR.
Table of Contents
- What counts as a weasel word in a 10-K
- Why hedging is everywhere in annual reports
- Examples of wording that limits a claim
- What research associates with hedged and obfuscated reporting
- A method for assessing hedging in an annual report
- Plain-English benchmarks as a reference point
- Moving from reading to systematic measurement
- Perspective
- Applying this analysis with Lacuna Index
- FAQ
- Sources
What counts as a weasel word in a 10-K
In the linguistics literature the phenomenon is called hedging. A 2009 AMCIS paper by Humpherys and colleagues cites Lakoff's (1972) definition of hedges as words whose job is to make things more or less fuzzy, and Hyland's (1998) description of them as devices such as perhaps, might and maybe that signal the writer's degree of confidence in a statement. "Weasel word" is the informal label for the same behavior when it is used to avoid commitment.
The same paper catalogs the vocabulary, which gives a reader a working lexicon:
| Class | Examples |
|---|---|
| Epistemic modal verbs | might, may, could, would, should, ought |
| Hedging adjectives | about, apparent, approximate, many, most, possible, potential, probable, speculative |
| Hedging adverbs | almost, approximately, essentially, generally, likely, mostly, normally, often, partially, possibly, presumably, relatively, somewhat, usually, virtually |
| Hedging nouns | possibility, probability |
A second group operates at the level of the sentence rather than the word. ICAEW's 2026 article on auditor vocabulary, drawing on Dr Koller's study, distinguishes pervasive hedging (qualifiers tied to likelihood or to conditions, often stacked) from modifiers that make a statement less than absolute or frame it as opinion: "we believe," "we consider," "in our view." Multiple qualifiers in close proximity can give an impression of high subjectivity or uncertainty.
Why hedging is everywhere in annual reports
A reader who flags every "may" will flag every page. The Humpherys paper notes that 10-K writers deliberately mark the uncertain nature of their discussion, typically with a Safe Harbor section listing terms such as believes, expects, may and should. It adds that the practice is encouraged to promote transparency under section 27A of the Securities Act of 1933, section 21E of the Securities Exchange Act of 1934 and the Private Securities Litigation Reform Act of 1995. The presence of hedging therefore does not by itself signify deception.
ICAEW makes the same point from the audit side: many qualifiers accurately represent uncertainties, because uncertainty is inherent in business. The concern is pervasive qualification that produces unnecessarily complex sentences and invites the question of whether each qualifier is truly necessary.
The analytical distinction therefore runs between two functions:
- Calibrating hedges attach to a genuinely uncertain event and are accompanied by the facts that bound the uncertainty (a range, a driver, a date, a named counterparty).
- Evasive hedges attach to a matter management could state definitively, or substitute for the supporting detail a claim would normally carry.
Examples of wording that limits a claim
The pattern is easiest to see in a before-and-after format. The following constructed illustrations are not quotations from any filing; they show the mechanics.
| Qualified wording | What the qualifier does | A verifiable alternative |
|---|---|---|
| "We believe our controls are generally effective." | Cognitive modifier plus a degree adverb; no clear conclusion. | A direct statement that disclosure controls were effective or not effective. |
| "Demand in international markets may continue to improve." | Modal verb, vague geography, vague direction. | Named regions with reported revenue and growth rates. |
| "We expect to realize substantial synergies over time." | Unquantified magnitude and unbounded timing. | A dollar target, a realization schedule, and disclosed progress to date. |
| "Key offerings performed relatively well." | Unnamed product, comparative with no baseline. | Product-level or segment-level results against the prior period. |
One published instance concerns controls disclosures. Deloitte's DART guidance records that SEC staff expect management to state without qualifying language whether disclosure controls and procedures are effective or ineffective, and that phrases such as "adequate," "effective except for" or "reasonably effective" are unacceptable. The regulator, in other words, treats certain hedges as non-answers.
Executive letters supply a second case. According to Institutional Investor's 2015 reporting, Laura Rittenhouse of Rittenhouse Rankings scores executive communication with a method she calls FOG (fact-deficient, obfuscating generalities). She rated WellPoint CEO Joseph Swedish's 2013 year-end letter "circular," meaning it supported a statement by repeating it in different or stronger terms, while General Motors CEO Mary Barra's 2013 letter earned high marks for candor because it explained what the balance-sheet strengthening and strategic investments actually were. The contrast is instructive: the difference was less about the count of hedges than about whether the claim was followed by substance.
What research associates with hedged and obfuscated reporting
Academic work links textual features of 10-Ks to outcomes, with the caveat that each study measures aggregate tendencies and not the intent of any single filer.
- Fraud samples. Moffitt and Burns examined 202 fraudulent and non-fraudulent 10-Ks across 25 linguistic cues and found that authors of fraudulent filings used more complex words, more words of achievement and cause, and more qualifying conjunctions, while truthful filings used more present-tense verbs and read more easily on the Flesch Reading Ease measure. A 2026 study of Indonesian public companies reported significantly higher hedging-language use in fraudulent reports (p=0.033), a different jurisdiction and a reminder that findings are sample-specific.
- Cost of capital. A study in the Journal of Financial and Quantitative Analysis found that firms with larger 10-K files and more uncertain and weak modal words had stricter loan terms and greater future stock-price crash risk. A 2021 Journal of Corporate Finance study of U.S. public firms from 1995 to 2017 associated greater textual complexity with a higher cost of equity, about 54 basis points when moving from the first to the third quartile.
- Interpretation. Cassidy's 2015 University of Georgia thesis observes that a high amount of hedging in an annual report could indicate that managers are very uncertain about the future. That is one reading. The same density can reflect genuine uncertainty, conservative legal drafting or deliberate vagueness, which is why the language signal must be tested against records.
A method for assessing hedging in an annual report
The approach below treats hedging as a flag that opens an inquiry and never as a verdict.
Step 1: Locate the qualified claims that matter
Not every sentence carries equal weight. Prioritize claims about performance drivers, strategic commitments, controls, customers, margins and guidance, since these are the statements an investor or governance reviewer would rely on. Boilerplate in the Safe Harbor section is expected and carries little analytical information.
Step 2: Classify the qualifier
Use the lexicon above to mark modal verbs, hedging adverbs, cognitive modifiers and stacked combinations. A sentence with three layers (for example, "we believe we may generally be able to") warrants closer reading than one with a single modal verb tied to a named risk.
Step 3: Check whether the claim has an anchor
Ask five questions of each flagged sentence:
- Is the subject named (a product, jurisdiction, statute, counterparty) or generic?
- Is the magnitude quantified, and against what baseline?
- Is the time frame stated?
- Does the next sentence supply the supporting fact, or does it restate the claim in broader terms?
- Does the statement repeat from year to year with the same wording while results change?
Step 4: Compare against the records
The SEC's investor bulletin on reading a 10-K directs readers to risk factors, unresolved staff comments, MD&A, audited financial statements and notes, the auditor's report, controls disclosures and non-GAAP measures, all of which are available on EDGAR. The same bulletin notes that laws and regulations prohibit materially false or misleading statements in 10-Ks and that the Sarbanes-Oxley Act requires the CEO and CFO to certify the filing's accuracy. A hedged narrative claim can be tested against these sections: a statement that "we believe margins may improve" is checkable against segment results, cost disclosures and the non-GAAP reconciliation.

Step 5: Apply the balance test
DLA Piper's guidance on annual shareholder letters recalls that SEC guidance on MD&A calls for a balanced view of the underlying dynamics of the business, including instances when the company failed to realize goals, if material. It also warns that excessively promotional language can obfuscate or deemphasize important information. A narrative that qualifies every risk and every shortfall while stating successes without qualification fails the balance test even if each individual hedge is defensible.
Plain-English benchmarks as a reference point
The SEC's Plain English Handbook gives a regulatory yardstick for what clear disclosure looks like. It describes plain English as using words economically and at a level the audience can understand, with tight sentence structure and a direct tone, and notes that Rule 421(b) calls for prospectuses that avoid vague boilerplate, legalistic presentations and repetition. The rule applies to prospectuses, not to 10-Ks as such, but its criteria translate into a practical reading standard: short sentences, specific nouns, no reliance on defined terms to hide meaning.

Moving from reading to systematic measurement
Manual review works for a single filing. Coverage across a sector, and across years, requires a consistent process. Several related analyses elaborate individual steps:
- Hedging language detection in corporate disclosures covers how hedging is operationalized in analysis.
- Tracing MD&A claims to GAAP describes how narrative statements are tied to reported figures.
- Monitoring disclosure consistency across filings addresses year-over-year drift in wording.
- Deception language in earnings calls extends the same lens to spoken disclosure.
Perspective
Language is the cheapest part of a disclosure to produce and the most expensive to verify, which is why the aspiration-to-execution gap tends to open first in wording. Our view is that a hedge should never be scored in isolation. A qualifier is neutral until it is set beside what the company later delivered.
For that reason the Lacuna methodology works only from public records, applies one standard across every covered issuer, and publishes governance logs for all scoring, so that each score can be traced to the passage and the record behind it. Counting "may" and "could" produces a readability statistic. Comparing a narrative claim with delivery produces an execution score that an analyst, a proxy advisor or a journalist can audit.
The practical consequence is that cohort comparison matters as much as single-filing reading. A hedging density that looks alarming in one issuer may be ordinary for its sector, and one that looks ordinary may be an outlier. Sector benchmarks supply that denominator.
Glen
Applying this analysis with Lacuna Index
Lacuna Index quantifies the gap between what public companies claim and what they deliver, using only public records such as SEC filings, earnings calls, proxy statements and press releases. Sector benchmarks are free to browse, for example Financials, Health Care and Industrials. Company-level forensic reports, with execution and narrative scores, require a subscription.

The Cohort Pulse and Sector Sweep plans are described on the pricing page, and the methodology walkthrough explains how scoring is built and logged.
FAQ
Are weasel words in an annual report always a red flag?
No. U.S. securities law encourages forward-looking statements accompanied by cautionary language, so hedges such as "may" and "expects" are routine and expected in a 10-K. The flag is raised when qualifiers are stacked, attached to matters management could state definitively, or detached from the supporting facts.
Which hedging words appear most often in annual reports?
Modal verbs (may, might, could, should, would) lead the list, followed by adverbs such as generally, approximately, relatively, usually and potentially, and by cognitive phrases such as "we believe" and "we expect." The Humpherys paper cited above provides the fuller lists of adjectives, adverbs and nouns.
Can hedging language predict fraud?
It can contribute to a pattern, but it does not establish fraud. Studies of 10-K samples have found differences in language between fraudulent and non-fraudulent filings, including more qualifying conjunctions and more complex words, and a 2026 Indonesian study found higher hedging in fraudulent reports. Results depend on the sample, and any flag requires confirmation against financial records.
How can a reader test whether a hedged claim is substantiated?
Place the claim beside the records that bear on it: MD&A, audited statements and notes, the auditor's report, controls disclosures and non-GAAP reconciliations, all on EDGAR. Then compare the claim with subsequent filings to see whether delivery matched the stated expectation.
Sources
- Humpherys et al., hedging and fraudulent financial statements (AMCIS 2009)
- Moffitt and Burns, investigating obfuscation in 10-Ks (AMCIS 2009)
- ICAEW, Mind your language: inside the auditor's vocabulary
- Deloitte DART, Disclosure Controls and Procedures
- Institutional Investor, How to Tell if a CEO Is Lying
- Annual Report Readability, Tone Ambiguity, and the Cost of Borrowing
- Annual report readability and the cost of equity capital
- Cassidy, Between the Hedges (University of Georgia, 2015)
- JATIT, fraud detection in Indonesian public companies (2026)
- SEC, Investor Bulletin: How to Read a 10-K
- SEC, A Plain English Handbook
- DLA Piper, Annual shareholder letter considerations and guidance
