A small set of behaviors account for a disproportionate share of unnecessary negative sentiment
Picture a Fortune 500 CFO 45 minutes before a quarterly earnings call. The prepared remarks are tightly rehearsed. Legal has blessed every sentence. The IR team has spent four weeks stress-testing disclosures. Then Q&A begins and all that discipline evaporates.
The analyst on the first question asks whether the company is seeing ‘meaningful deterioration in enterprise demand’. The CFO pauses, then answers: ‘Yes, we are seeing some deterioration in enterprise demand, particularly as customers become more cautious in this environment.’
It’s a reasonable answer. It’s accurate. In the boardroom’s mental model, it’s a moment of transparency that builds credibility with the Street.
In the data centers running the trading algorithms, it’s something else entirely. Within seconds of that exchange, transcript-parsing engines operated investment funds, NLP-based sentiment platforms and AI-driven trading systems have flagged the cluster: ‘deterioration’, ‘cautious’ ‘environment’ – three high-weight negative constructs in a single answer. The position is already being marked. By the time the call ends, the stock has given back a third of what the EPS beat had delivered.
This is not a hypothetical. It is the architecture of a problem playing out across earnings seasons, quarter after quarter, at companies large and small.
As we documented in our previous article, ‘Why stocks fall after earnings: The hidden power of tone, language and sentiment’, more than half of US equities trade lower within 72 hours of an earnings release, and roughly 45 to 50 percent of EPS beaters still trade down the next day. That shift is about how management teams sound, and the Q&A is where the sound becomes a signal.
What the research now confirms
In a 2025 study, the Financial Review introduced the concept of ‘tone distance’: the degree of variance between different executives’ language on the same call. It concluded that greater tone distance is independently associated with lower cumulative abnormal returns around earnings announcements, even after controlling for earnings surprise and general tone levels. Inconsistency between what the CEO and CFO say, in other words, is itself a priced signal.
That finding sits within a broader body of converging evidence. Researchers at NYU demonstrated that NLP-derived sentiment from earnings transcripts is significantly correlated with subsequent stock and bond returns after controlling for earnings surprise, an effect that persists across all sectors. A University of Tilburgstudy using GPT-4o confirmed the relationship holds after controlling for both earnings surprise and firm size. Stanford’s Graduate School of Business published research showing that executive evasiveness during Q&A measured by machine learning against nearly 1,800 manager responses predicts future earnings misses and lower stock returns. A trading strategy built on evasiveness scores yields positive risk-adjusted returns. These findings show that Q&A is not merely a conversation, but a data feed. And it is the least prepared part of the event.
The preparation gap is structural. Roughly 22 percent of public companies conduct no formal Q&A rehearsal, while nearly half spend two hours or less preparing for a segment that often represents more than half of an earnings call and can influence investor perception more than the prepared remarks. At the same time, a growing number of analysts and investors are consuming earnings materials through AI-generated summaries that emphasize opening statements and compress nuance. As a result, the first sentence of a Q&A response may now carry more weight than ever, often becoming the only sentence a decision-maker remembers.
Across earnings calls, a small set of recurring communication behaviors account for a disproportionate share of unnecessary negative sentiment. None of these involves disclosure violations. All feel instinctive to the executives deploying them.
- Leading with the problem. When asked about margin pressure, executives confirm it first: ‘Margins were definitely pressured during the quarter.’ When asked about demand: ‘We did see softness there.’ These confirmations feel like transparency, but placing the negative framing at the beginning of a response disproportionately anchors how the entire answer is interpreted by both sentiment models and investors.
- Mirroring the analyst’s language. Analysts construct questions using bearish architecture by design. When management mirrors that phrasing in the response as illustrated earlier with the response ‘Yes, we are seeing deterioration in enterprise demand, particularly as customers become more cautious’, the transcript has now twice reinforced each negative construct. Management teams should break the mirroring: acknowledge the topic, answer directly and reframe in language management controls.
- Repetition across the full call. Executives often describe the same operational challenge using slightly different phrasing across multiple answers: ‘revenue declined’, ‘growth slowed’, ‘demand weakened’, ‘visibility remains limited’. Each statement is accurate. In aggregate, they cluster into a sentiment signature that disproportionately shapes how both algorithms and investors interpret the call as a whole. Investors don’t remember every metric. They remember what management emphasized most, repeated most often and framed with the greatest conviction.
- Ending on unresolved uncertainty. Responses that close with ‘we’ll have to wait and see’, ‘visibility remains limited’ or ‘the environment remains uncertain’ feel prudent. In standard financial NLP lexicons, including the Loughran-McDonald dictionary used by many institutional sentiment models, uncertainty language is among the highest-weight negative categories in financial text. More effective communicators end answers by returning to execution, preparedness and operational control, leaving the tone of the response on solid ground rather than suspended.
The table below illustrates the difference between traditional Q&A responses and more effectively structured alternatives. Every revised response communicates the same operational reality as the original. The difference is in sequencing, linguistic architecture and emphasis.
| Traditional response | Structured narrative framework | Why it works better |
| ‘Demand weakened due to macro pressure and customer caution.’ | ‘While customers remained selective, demand across our core business stayed resilient and we continued executing on our strategic priorities.’ | Leads with resilience before contextualizing the pressure. |
| ‘We remain cautious given ongoing uncertainty.’ | ‘We continue managing the business with flexibility and discipline while staying focused on execution.’ | Replaces passive uncertainty with action-oriented positioning. |
| ‘Margins were pressured by freight costs and lower utilization.’ | ‘Margins reflected temporary freight and utilization headwinds, while operational efficiency initiatives remained firmly on track.’ | Pairs the challenge directly with mitigation. |
| ‘Tariff uncertainty is creating pressure.’ | ‘We continue navigating evolving tariff dynamics while leveraging operational flexibility and supply-chain diversification.’ | Pairs the challenge directly with management action. |
| ‘We are disappointed with the quarter.’ | ‘The quarter developed differently than expected, though several strategic initiatives continued progressing well.’ | Removes emotionally negative language while maintaining transparency. |
What disciplined preparation looks like and why it matters
Recognizing these patterns is not the same as solving them. The communication behaviors above are deeply instinctive. They developed in a market that rewarded them. Changing them under pressure requires structured rehearsal, not just awareness.
- Build the question universe early. Two weeks out, IR teams should construct a comprehensive list of adversarial questions – not just obvious ones, but those that probe the most challenging metrics and mirror the bearish language structures analysts are likely to deploy. Identifying them early creates the runway for structured response development rather than reactive preparation.
- Develop response frameworks, not scripts. Scripted answers are brittle and sound scripted. The goal is consistent architecture: acknowledge the question directly, answer it factually, introduce mitigating context and close with execution or preparedness language. This structure should be consistent across all negatively framed questions, regardless of topic.
- Rehearse out loud, under pressure. Frameworks exist on paper; instincts exist in the body. The only preparation method that reliably changes real-time language behavior is live verbal rehearsal under conditions that simulate call pressure. IR team members should organize a ‘murder board’ exercise ahead of every earnings call, acting as analysts and presenting adversarial questions to management until the structured response becomes instinctive. Research on executive communication consistently shows that hesitation, vocal strain and long pauses during Q&A are themselves interpreted as negative signals.
- Calibrate repetition across the full call. Before the call, review the complete set of anticipated Q&A responses and flag how often specific negative constructs appear in aggregate. If ‘pressure’, ‘softness’ or ‘uncertainty’ surface repeatedly across multiple answers, that pattern will accumulate in the transcript’s overall sentiment signature. A deliberate calibration pass should ensure negative constructs appear proportionally alongside positive or neutral framing, which should improve how the full call scores.
There is a version of this argument that management teams sometimes hear as ‘you need to spin your results more carefully’. That is not the argument. Spin involves misrepresentation. Every framework in the table above discloses the same facts as the traditional response it replaces. The difference is that earnings communication has always been a discipline, and that discipline has changed. For decades, it was primarily about legal compliance and messaging consistency. Today, it also requires understanding how machine-readable systems interpret linguistic patterns.
Companies that achieve genuine operational performance and then communicate it in ways that machines systematically score as uncertain, defensive or deteriorating are not experiencing a market failure. They are experiencing a communication failure, one that is increasingly preventable.
In today’s market, Q&A is no longer simply a conversation. It is structured market communication consumed simultaneously by humans and machines. The management teams best positioned in this environment are those who have closed the preparation gap: who treat the Q&A session with the same rigor they apply to the financial statements themselves.
Dan Joldzic is CEO and Naya Bermudez is director, investor relations products at Alexandria Technology, an AI and natural language processing firm that offers sophisticated text and sentiment analysis for investors and public companies.
