Dave recently shared an academic paper arguing that AI is not just analyzing MD&As — but actually shaping how they’re written. The author, Professor Keren Bar-Hava of the Hebrew University of Jerusalem, just released a new paper summarized in this CLS Blue Sky Blog, suggesting two empirical tools to measure how companies tailor their MD&As to algorithms and when AI-induced disclosure pressure has caused tone to cross the line from optimism to manipulation.

The first tool, the AI Orientation Score, ranges from 0 to 5 and assesses how machine-optimized a disclosure is based on measurable language, keyword usage, structured formatting, impersonal tone, and tonal consistency. A higher score suggests that the MD&A was likely crafted with machine readers in mind.

The second tool, the AI Manipulation Exposure Index (AI-MEX), also scored from 0 to 5, captures rhetorical red flags that may signal tone manipulation. These include upbeat language despite poor financials, absence of key performance indicators, vague aspirational claims, and excessive repetition of positive terms.

Professor Bar-Hava applied these tools to analyze 80 MD&As from 20 large S&P 100 firms between 2021 and 2024 (using both ChatGPT and Gemini, which arrived at nearly identical conclusions) and found that:

AI Orientation Scores rose steadily, indicating growing use of algorithm-friendly narrative techniques in MD&A disclosures.

AI-MEX scores were significantly higher for firms with weak fundamentals, such as negative net income or deteriorating cash flow.

By 2024, over 60 percent of companies scored high on both indices, signaling a dual strategy of structural optimization and tone management.

She concludes that these trends are . . . not great (my words, not hers). She worries that they will erode investor trust and confidence in this important narrative disclosure that is often the “only section where strategy, performance, and uncertainty are discussed together in plain language.” She suggests that anyone on the reviewing end of MD&A start to incorporate these diagnostic tools to help understand when the narrative doesn’t align with the numbers and holding companies “accountable for how tone, metrics, and structure align.”

How should public companies and their lawyers be receiving and responding to this?

First, be aware that AI is one of the several audiences of your disclosures, and that your disclosures may be screened or assessed by an algorithm. On RealTransparentDisclosure.com, Broc recently shared some great practice pointers for drafting with that in mind. (The last one in particular jumped out at me: “Recognize that today’s disclosures train tomorrow’s AI models. The tone and style choices you make now can set future industry expectations—so lead wisely.”)

Second (and this aligns with Broc’s point no. 4), ensure you’re still presenting an accurate, fair and balanced picture of company performance. Investors may start considering to what extent you’ve machine-optimized disclosure and when and whether it starts to border on manipulation (or even misrepresentation) by applying these or other analysis tools. To that end, companies may also want to start considering how their disclosures score on AI Orientation and AI-MEX.

I’d also argue that outside counsel (through human review) may be well-positioned to help clients avoid high AI-MEX scores by looking at all the disclosures holistically and asking questions where disclosures (including tone) appear inconsistent. That can be a hard task for the preparer who has been living and breathing the numbers and disclosures for a few weeks, while a more outside observer well-versed on the company and its industry reading a full draft for the first time might immediately pick up on language that comes off as overly optimistic or inconsistent.

It goes without saying that “the company and its management are responsible for the accuracy and adequacy of their disclosures notwithstanding. . . any action or absence of action” by company or investor AI. (See what I did there?)

– Meredith Ervine