{"id":1292,"date":"2026-11-14T04:01:00","date_gmt":"2026-11-14T09:01:00","guid":{"rendered":"https:\/\/drugchatter.com\/insights\/?p=1292"},"modified":"2026-11-04T21:23:27","modified_gmt":"2026-11-05T02:23:27","slug":"before-the-next-earnings-call-ask-what-does-ai-say-about-us","status":"publish","type":"post","link":"https:\/\/drugchatter.com\/insights\/before-the-next-earnings-call-ask-what-does-ai-say-about-us\/","title":{"rendered":"Before the Next Earnings Call, Ask: What Does AI Say About Us?"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

Pfizer held its third-quarter 2026 earnings call on Tuesday, November 3, at 10:00 a.m. EST. The company said it would post the transcript and webcast replay within 24 hours and keep them online for at least 90 days. [17] Any analyst, journalist, or retail investor can now paste that transcript into a chatbot and ask for a summary. Most companies have not checked what the chatbot would say.<\/p>\n\n\n\n

This article covers one question: what does an AI system say about your company, your drugs, and your guidance, and who inside the company owns the answer? We did not run a new multi-model test for this piece. We reconstruct what public studies, court filings, regulatory text, and vendor indices show, calculate a few derived figures from them, and propose a pre-call audit that Medical Affairs, Regulatory, Pharmacovigilance, Investor Relations, Legal, and Brand teams can run. Original frameworks are labeled as ours. Calculations are labeled as analysis.<\/p>\n\n\n\n

The Short Answer<\/strong><\/h2>\n\n\n\n

Run a pre-call AI audit. Ask the major AI assistants the questions an investor, analyst, patient, or reporter would ask. Compare each answer with five sources: the current prescribing information, the latest 8-K and 10-Q, the press release, the call script, and ClinicalTrials.gov. Record each discrepancy, rank it by severity, and send it to the function that owns the source document. Do this before the call, and again one day after.<\/p>\n\n\n\n

The reasons are documented. About seven in ten institutional investors in a Brunswick Group survey say AI has changed how they approach earnings calls, according to Axios. [6] In a Pew Research Center study of 900 US adults, users clicked a traditional result link in 8% of visits that showed a Google AI summary and in 15% of visits that did not. [1] In a BMJ Quality & Safety study, 26.0% of 484 scored chatbot answers about common drugs had accuracy issues against the reference. [10] None of these studies measures your company. That is the point of the audit.<\/p>\n\n\n\n

Why Do Earnings Calls Now Have an AI Audience?<\/strong><\/h2>\n\n\n\n

Earnings calls now have two audiences. People listen live. Software reads the replay, the transcript, the 8-K, and the press release, and then answers questions for people who did not listen.<\/p>\n\n\n\n

What Do Investors Say About Using AI Around Earnings Calls?<\/strong><\/h3>\n\n\n\n

Axios reported on a Brunswick Group study of institutional investors in February 2026. It said 54% of respondents call AI outputs an important part of their investment research. About 7 in 10 said AI has changed how they approach earnings calls. [6]<\/p>\n\n\n\n

Axios also reported that 46% of respondents are more likely to skip earnings calls and review AI-generated summaries instead. Four in 10 said they trust AI summaries of financial content as much as those written by sell-side analysts. [6] The Axios text we reviewed does not state the sample size or the survey dates, so treat the percentages as directional.<\/p>\n\n\n\n

\n

46% of surveyed institutional investors say they are more likely to skip earnings calls and read AI-generated summaries instead (Brunswick Group survey, as reported by Axios, February 5, 2026). [6]<\/p>\n<\/blockquote>\n\n\n\n

Retail investors show a similar pattern. Investing.com surveyed 938 US investors. ChatGPT was the most common tool, used for investing decisions by 54% of respondents, and 39% worried that AI tools can get things wrong. [19] The report states a headline figure of 65% of investors using AI, while its own breakdown (24% regularly, 27% occasionally, 11% once or twice) adds up to 62%. [19] We cite the breakdown and flag the mismatch.<\/p>\n\n\n\n

What the Investor Surveys Show<\/strong><\/h4>\n\n\n\n

Both surveys point the same way. Investors use AI to read financial content, and some investors use AI summaries in place of the primary event. [6][19]<\/p>\n\n\n\n

What the Investor Surveys Do Not Show<\/strong><\/h4>\n\n\n\n

Neither survey tells you which AI answers are wrong about a specific issuer. Neither separates pharma from other sectors. Neither measures whether an AI summary changed a trade. Those gaps are why a company-level audit has value.<\/p>\n\n\n\n

How Many Patients Ask AI About Drugs and Health?<\/strong><\/h3>\n\n\n\n

OpenAI reported in January 2026 that more than 40 million people submit health-related prompts to ChatGPT each day, and that health makes up more than 5% of all ChatGPT messages worldwide. [12] About one in four of its more than 800 million regular users submits at least one health prompt per week. [12] These are OpenAI’s own figures, reported through Medical Economics. We could not access the original 18-page report, so we cite the secondary summary. [12]<\/p>\n\n\n\n

An earnings call touches this audience indirectly. A patient who reads an AI answer about your drug’s side effects, price, or availability is part of the same information system as the analyst who reads an AI answer about your pipeline. Both answers draw on public sources. Both can be wrong in ways your team can check.<\/p>\n\n\n\n

Do AI Summaries Replace the Click to Your Own Source?<\/strong><\/h3>\n\n\n\n

Pew Research Center tracked the browsing of 900 US adults who agreed to share their online activity in March 2025. [1] Users clicked a traditional search result link in 8% of visits when a Google AI summary appeared and in 15% of visits when none appeared. [1] Users clicked a link inside the AI summary in 1% of visits. [1] Users ended their browsing session after 26% of search pages with an AI summary and after 16% of pages without one. [1]<\/p>\n\n\n\n

Our calculations from those four figures, labeled as analysis:<\/p>\n\n\n\n

    \n
  • The click rate on traditional results was 46.7% lower in visits with an AI summary ((15 – 8) \/ 15).<\/li>\n\n\n\n
  • The session-end rate was 10 percentage points higher, a 62.5% relative increase ((26 – 16) \/ 16).<\/li>\n\n\n\n
  • Traditional result links drew eight times the clicks of links inside the summary (8% versus 1%).<\/li>\n\n\n\n
  • If an investor reads the AI summary and stops, your IR page, your 10-Q, and your label never enter the conversation.<\/li>\n<\/ul>\n\n\n\n

    Pew studied general Google searches. It did not study pharma, investor, or earnings queries. Apply the numbers as a general signal about click behavior, not as a pharma finding.<\/p>\n\n\n\n

    The Five Findings That Matter<\/strong><\/h2>\n\n\n\n

    These findings can be quoted on their own. Each carries its source.<\/p>\n\n\n\n

      \n
    1. Pfizer scheduled its Q3 2026 earnings call for November 3, 2026, with a transcript and replay promised within 24 hours and retained for at least 90 days. [17]<\/li>\n\n\n\n
    2. In the Pew study, clicks on traditional links fell from 15% to 8% of visits when a Google AI summary appeared. [1]<\/li>\n\n\n\n
    3. In the BMJ Quality & Safety study, 126 of 484 scored Bing Copilot answers (26.0%) had accuracy issues against the reference source. [10]<\/li>\n\n\n\n
    4. In the Long Island University study, ChatGPT gave satisfactory answers to 10 of 39 real drug-information questions (25.6%, our calculation). [11]<\/li>\n\n\n\n
    5. In the 5W Q2 2026 index, the top six pharma companies held 50.5% of estimated AI citation share across five AI engines (our sum of reported figures). [8]<\/li>\n<\/ol>\n\n\n\n

      What Does Regulation FD Say, and Does It Cover AI?<\/strong><\/h2>\n\n\n\n

      Regulation FD does not mention AI. It governs when an issuer or a person acting on its behalf discloses material nonpublic information to certain people. The rule text applies to AI-related situations only through its general structure, so we separate the rule text from our reading of it.<\/p>\n\n\n\n

      What Does Regulation FD Require?<\/strong><\/h3>\n\n\n\n

      Whenever an issuer, or any person acting on its behalf, discloses material nonpublic information, the issuer must make public disclosure of that information. [3] It must do so simultaneously when the disclosure is intentional and promptly when it is not. [3]<\/p>\n\n\n\n

      A disclosure is intentional when the person making it knows, or is reckless in not knowing, that the information is both material and nonpublic. [4] Promptly means as soon as reasonably practicable, and in no event after the later of 24 hours or the start of the next day’s trading on the New York Stock Exchange, after a senior official learns of a non-intentional disclosure. [4]<\/p>\n\n\n\n

      Which Recipients Does the Rule Name?<\/strong><\/h4>\n\n\n\n

      The general rule applies to disclosures to four groups: brokers or dealers and their associated persons, investment advisers, investment companies, and holders of the issuer’s securities where it is reasonably foreseeable that the holder will trade on the information. [3]<\/p>\n\n\n\n

      Which Disclosures Fall Outside the Rule?<\/strong><\/h4>\n\n\n\n

      The rule does not apply to disclosures to a person who owes the issuer a duty of trust or confidence, such as an attorney, investment banker, or accountant. [3] It also does not apply to disclosures to a person who expressly agrees to keep the information in confidence. [3] Certain registered offering communications are also excepted. [3]<\/p>\n\n\n\n

      How Does Regulation FD Treat an Earnings Call Webcast?<\/strong><\/h3>\n\n\n\n

      An issuer meets the public disclosure requirement by furnishing or filing a Form 8-K. [4] It can instead use another method, or combination of methods, reasonably designed to provide broad, non-exclusionary distribution of the information to the public. [4] The Pfizer arrangement, a public webcast announced in advance with registration details posted from September 22, 2026, is an example of an announced public-access mechanism. [17] We do not offer a legal conclusion about any specific company’s compliance.<\/p>\n\n\n\n

      Can an Employee’s Chatbot Prompt Become a Selective Disclosure?<\/strong><\/h3>\n\n\n\n

      This section is our interpretation, not legal advice. Reg FD is triggered by disclosure to persons the rule names. [3] A chatbot vendor is not on that list. The risk we see sits elsewhere: an employee pastes draft guidance, a trial readout, or an unreleased label negotiation into a public AI tool before the call.<\/p>\n\n\n\n

      Whether that creates a Reg FD problem depends on facts a lawyer must assess, including who receives the information and under what confidentiality terms. [3] It can also create confidentiality and trade-secret problems that have nothing to do with Reg FD. A pre-call audit should therefore use only public inputs, and the audit protocol below says so.<\/p>\n\n\n\n

      What Happens If an AI Tool Summarizes the Call Incorrectly?<\/strong><\/h3>\n\n\n\n

      Reg FD has no provision that makes an issuer responsible for a third party’s AI summary. Other law may apply to the issuer’s own statements. The Moffatt v. Air Canada decision, discussed below, shows a tribunal holding a company responsible for what its own chatbot said. [14] That case concerned the company’s own chatbot, not a third-party tool. Do not stretch it.<\/p>\n\n\n\n

      What Happens to Forward-Looking Statements Inside an AI Summary?<\/strong><\/h2>\n\n\n\n

      Forward-looking statements are the center of a pharma earnings call: launch timing, peak sales, trial readouts, regulatory submissions, and guidance. The statutory safe harbor protects a forward-looking statement only if it is identified as forward-looking and accompanied by meaningful cautionary statements, or if the plaintiff cannot prove actual knowledge that the statement was false or misleading. [5]<\/p>\n\n\n\n

      What Does the Safe Harbor Require for Oral Statements?<\/strong><\/h3>\n\n\n\n

      For oral statements, the cautionary requirement is treated as met when the speaker says the statement is forward-looking and that actual results could differ materially. The speaker must also point listeners to a readily available written document that contains the additional cautionary factors, and must identify that document. [5] That is why every earnings call opens with a forward-looking statements preface.<\/p>\n\n\n\n

      Where Can the Caveat Disappear?<\/strong><\/h3>\n\n\n\n

      This is our analysis. An AI summary of a transcript can compress a sentence such as “we expect to file in the second half” into “the company will file in the second half”. A summarizer can drop the preface that supplied the caution.<\/p>\n\n\n\n

      We do not claim a court would treat the summary as the company’s statement. The safe harbor attaches to the statement as the company made it. [5] The operational risk is that an investor acts on the compressed version and the company learns about it later. We classify this as a Type 2 error in the taxonomy below.<\/p>\n\n\n\n

      What to Check Before the Call<\/strong><\/h4>\n\n\n\n

      Ask each AI tool to summarize last quarter’s transcript. Check whether the summary keeps the timing qualifiers, the risk-factor reference, and the difference between guidance and aspiration.<\/p>\n\n\n\n

      What Have AI Systems Gotten Wrong About Drugs and Companies? Documented Cases<\/strong><\/h2>\n\n\n\n

      The cases below are public. None of them measures the current accuracy of any model, because models change. Each one shows a type of error.<\/p>\n\n\n\n

      What Did the BMJ Quality & Safety Study Find About Chatbot Drug Answers?<\/strong><\/h3>\n\n\n\n

      Andrikyan and colleagues asked Bing Copilot 10 patient questions about each of the 50 most prescribed US outpatient drugs of 2020, producing 500 answers. [10] The paper appeared in BMJ Quality & Safety (2025;34:100-109) and was published online on October 1, 2024. [10]<\/p>\n\n\n\n

      How Was the Study Run?<\/strong><\/h4>\n\n\n\n

      The team queried the chatbot in English in “balanced” mode through a New York VPN in April 2023. [10] The reference source was drugs.com. [10]<\/p>\n\n\n\n

      What Did the Study Report?<\/strong><\/h4>\n\n\n\n

      Mean accuracy was 88.7% and mean completeness was 76.7%. [10] Accuracy issues appeared in 126 of 484 answers (26.0%), and 16 answers (3.3%) were fully inconsistent with the reference. [10] The mean Flesch Reading Ease score was 37.2, which the authors describe as difficult to read. [10]<\/p>\n\n\n\n

      In a non-random subset of 20 answers, experts judged 66% potentially harmful and 22% likely to cause death or severe harm. [10] The authors selected that subset for low accuracy, low completeness, or safety risk. [10] Do not apply 66% or 22% to all chatbot answers. The paper does not support that use.<\/p>\n\n\n\n

      What Are the Study’s Limits?<\/strong><\/h4>\n\n\n\n

      The authors list several limits. They tested one chatbot, one language, one location, and one month. Chatbot output may have changed since April 2023. Large language model output is stochastic. They did not simulate follow-up dialogue. [10]<\/p>\n\n\n\n

      What Did the Long Island University Study Find About ChatGPT and Drug Questions?<\/strong><\/h3>\n\n\n\n

      Pharmacists at Long Island University posed 39 questions to the free version of ChatGPT in May 2023. The questions came from 45 real questions sent to the university’s drug information service, and six were excluded because no literature was available. [11] ChatGPT gave satisfactory answers to 10 of 39. [11] Of the 29 unsatisfactory answers, 11 did not directly address the question, 10 were inaccurate, and 12 were wrong or incomplete, with overlap among categories. [11]<\/p>\n\n\n\n

      Our calculation: 10 of 39 is 25.6% satisfactory and 29 of 39 is 74.4% unsatisfactory. The sample is small and the model is old. Use the study as evidence that real drug-information questions expose errors, not as a current accuracy rate.<\/p>\n\n\n\n

      What Did the Guardian Find About Google AI Overviews and Health Queries?<\/strong><\/h3>\n\n\n\n

      Euronews reported on January 12, 2026 that a Guardian investigation found inaccurate health information in some Google AI Overviews. [13] For liver blood test queries, the top result was a featured snippet pulled from an Indian hospital chain’s page, with many numbers and little context on age, sex, ethnicity, or nationality. [13] Google reportedly advised patients with pancreatic cancer to avoid high-fat foods, which experts told the Guardian could raise the risk of death. [13]<\/p>\n\n\n\n

      Euronews reported that Google had removed some health-related AI Overviews. It also said Google had not replied to its question when the article was published. [13] The liver-test result was a featured snippet, not an AI Overview, so the episode also shows that the line between “search result” and “AI answer” is not clean. [13]<\/p>\n\n\n\n

      What Does Wolf River Electric v. Google Show About Company-Level Errors?<\/strong><\/h3>\n\n\n\n

      Minnesota Lawyer reported in June 2025 that Wolf River Electric and four officers sued Google in Ramsey County District Court in March 2025, and that the case was removed to federal court. [15] The complaint alleges that a Google AI Overview said the company faced a lawsuit from the Minnesota Attorney General, and that the company says this is false. [15] It says several customers ended relationships in March 2025, with some contracts worth as much as $150,000. [15]<\/p>\n\n\n\n

      These are allegations from a complaint. The article we reviewed does not report an outcome. The case matters here because the alleged harm came from an AI answer about a company, not about a drug. [15]<\/p>\n\n\n\n

      What Did Moffatt v. Air Canada Decide About Chatbot Statements?<\/strong><\/h3>\n\n\n\n

      In Moffatt v. Air Canada, 2024 BCCRT 149, a British Columbia tribunal considered a chatbot that told a customer he could apply retroactively for a bereavement fare, while another page on the airline’s website said otherwise. [14] The tribunal rejected Air Canada’s argument that the chatbot was a separate entity. It held that the chatbot is part of the website and that the airline is responsible for all information on it. [14] It called the airline’s position “a remarkable submission”. [14]<\/p>\n\n\n\n

      The case concerns a company’s own chatbot. It does not decide who is liable for a third-party model’s answer. It does show that a tribunal will not accept “the bot said it” as a defense when the bot is yours. If your company deploys a medical-information or patient-support chatbot, that point applies to you.<\/p>\n\n\n\n

      Do Regulators Treat AI Claims About a Company as an Enforcement Issue?<\/strong><\/h3>\n\n\n\n

      On March 18, 2024, the SEC announced settled charges against Delphia (USA) Inc. ($225,000) and Global Predictions Inc. ($175,000), a combined $400,000, over statements about their use of artificial intelligence. [2] The SEC’s enforcement director at the time said firms that claim to use AI must ensure their representations are not false or misleading. [2] Those cases concern a company’s claims about its own AI use, not third-party AI answers about the company. They show that the SEC reads AI-related statements the way it reads other statements. [2]<\/p>\n\n\n\n

      Which Pharma Companies Do AI Engines Mention Most? The 5W Index<\/strong><\/h2>\n\n\n\n

      5W AI Communications, a PR firm, published a Pharma and Rx AI Visibility Index for Q2 2026. It tracked more than 60 patient and consumer prompts across six sub-categories on ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. Each prompt ran five times per engine in clean sessions. [8] Citation share is the firm’s estimate of the proportion of company-specific citations across those prompts. [8]<\/p>\n\n\n\n

      What Is the Citation Share of the Top 15 Pharma Companies?<\/strong><\/h3>\n\n\n\n
      Rank<\/th>Company<\/th>Estimated AI citation share, Q2 2026<\/th>Source<\/th><\/tr><\/thead>
      1<\/td>Eli Lilly<\/td>12.5%<\/td>[8]<\/td><\/tr>
      2<\/td>Novo Nordisk<\/td>11.5%<\/td>[8]<\/td><\/tr>
      3<\/td>Pfizer<\/td>8.5%<\/td>[8]<\/td><\/tr>
      4<\/td>Johnson & Johnson<\/td>7.0%<\/td>[8]<\/td><\/tr>
      5<\/td>Merck<\/td>6.0%<\/td>[8]<\/td><\/tr>
      6<\/td>AbbVie<\/td>5.0%<\/td>[8]<\/td><\/tr>
      7<\/td>AstraZeneca<\/td>4.0%<\/td>[8]<\/td><\/tr>
      8<\/td>Moderna<\/td>3.5%<\/td>[8]<\/td><\/tr>
      9<\/td>Amgen<\/td>3.0%<\/td>[8]<\/td><\/tr>
      10<\/td>Novartis<\/td>2.6%<\/td>[8]<\/td><\/tr>
      11<\/td>Bristol Myers Squibb<\/td>2.3%<\/td>[8]<\/td><\/tr>
      12<\/td>GSK<\/td>2.0%<\/td>[8]<\/td><\/tr>
      13<\/td>Sanofi<\/td>1.8%<\/td>[8]<\/td><\/tr>
      14<\/td>Gilead Sciences<\/td>1.5%<\/td>[8]<\/td><\/tr>
      15<\/td>Roche<\/td>1.3%<\/td>[8]<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      What Do the Numbers Say When You Add Them Up?<\/strong><\/h3>\n\n\n\n

      These sums are our analysis of the 5W figures.<\/p>\n\n\n\n

      Derived figure<\/th>Calculation<\/th>Result<\/th><\/tr><\/thead>
      Lilly plus Novo Nordisk share<\/td>12.5 + 11.5<\/td>24.0%<\/td><\/tr>
      Top 6 combined share<\/td>12.5 + 11.5 + 8.5 + 7.0 + 6.0 + 5.0<\/td>50.5%<\/td><\/tr>
      Top 10 combined share<\/td>Top 6 + 4.0 + 3.5 + 3.0 + 2.6<\/td>63.6%<\/td><\/tr>
      Top 15 combined share<\/td>Top 10 + 2.3 + 2.0 + 1.8 + 1.5 + 1.3<\/td>72.5%<\/td><\/tr>
      Lilly share relative to Merck<\/td>12.5 \/ 6.0<\/td>2.08 times<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      The pattern is concentration. Six companies hold half of the estimated citation share in the tracked prompts. A company ranked 11th or lower is mentioned far less often in the same prompts. That matters for an earnings call because a small-cap or mid-cap sponsor may find that AI tools describe its therapeutic area without naming it. We classify that as a Type 5 error below.<\/p>\n\n\n\n

      Does DTC Advertising Spend Predict AI Visibility?<\/strong><\/h3>\n\n\n\n

      The Fierce Pharma coverage of the 5W report cites iSpot.TV data: AbbVie and J&J spend $30 million to $50 million per month per drug on TV, and Lilly spends about $20 million per month on Zepbound. [9] Merck ranks fifth in citation share and AbbVie sixth. [8] We do not draw a causal link in either direction. [9] The data lets us say only that ad spend and citation share are different measures.<\/p>\n\n\n\n

      What Does the 5W Index Not Measure?<\/strong><\/h3>\n\n\n\n

      5W describes citation share as directional, not a precise market measure. It tracked one quarter. It publishes no per-engine or per-prompt results. Its publication date appears as both May and June 2026 on the page. It measures communications visibility, not safety, efficacy, or accuracy. [8] A company can be cited often and described wrongly. Visibility and accuracy are separate audits.<\/p>\n\n\n\n

      How Should You Classify AI Errors Before an Earnings Call? A Five-Type Taxonomy<\/strong><\/h2>\n\n\n\n

      The taxonomy below is our original framework, not an established industry classification. It sorts AI errors by the source document that would correct them, which tells you who owns the fix.<\/p>\n\n\n\n

      Type<\/th>Name<\/th>What it looks like<\/th>Authoritative source<\/th>Owner<\/th><\/tr><\/thead>
      1<\/td>Stale fact<\/td>The answer gives last year’s guidance, an old indication, or a withdrawn timeline.<\/td>Latest 8-K, 10-Q, press release<\/td>IR, Legal<\/td><\/tr>
      2<\/td>Stripped caveat<\/td>The answer turns a conditional statement into a firm one.<\/td>Call transcript and safe-harbor preface<\/td>IR, Legal<\/td><\/tr>
      3<\/td>Label drift<\/td>The answer misstates dosing, warnings, contraindications, or indication.<\/td>Current FDA prescribing information<\/td>Regulatory, Medical Affairs<\/td><\/tr>
      4<\/td>Misattribution<\/td>The answer credits your trial, safety signal, or product to another company, or the reverse.<\/td>ClinicalTrials.gov, label, filings<\/td>Medical Affairs, CI<\/td><\/tr>
      5<\/td>Absence<\/td>The answer describes your therapeutic area and does not mention your product or company.<\/td>Your public sources<\/td>Brand, Corporate Communications<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      What Is a Type 1 Stale Fact Error?<\/strong><\/h3>\n\n\n\n

      A Type 1 error states a fact that was true earlier. It is the most common failure we expect, because models and search indexes lag the filing date. The BMJ authors flag the same time problem for their own results: outputs from April 2023 may have changed. [10] Check dates in every answer.<\/p>\n\n\n\n

      How to Test for It<\/strong><\/h4>\n\n\n\n

      Ask for current-quarter guidance and compare the figure and date with the most recent 8-K Item 2.02 earnings release on EDGAR. [22]<\/p>\n\n\n\n

      What Is a Type 2 Stripped Caveat Error?<\/strong><\/h3>\n\n\n\n

      A Type 2 error removes the qualifier. The safe harbor for oral statements depends on the preface and the reference to a written document with risk factors. [5] A summary that deletes both has changed the evidence an investor sees.<\/p>\n\n\n\n

      How to Test for It<\/strong><\/h4>\n\n\n\n

      Feed a public transcript to each tool and ask for a one-paragraph summary. Mark each forward-looking sentence and check whether the summary keeps its qualifier.<\/p>\n\n\n\n

      What Is a Type 3 Label Drift Error?<\/strong><\/h3>\n\n\n\n

      A Type 3 error compares an AI statement about a drug with the current FDA label. The BMJ study and the LIU study both measured this kind of mismatch against a reference source. [10][11] The reference should be the current prescribing information, not a secondary database, because drugs.com served as the reference in the BMJ study and the authors note it lacked answers to some questions. [10]<\/p>\n\n\n\n

      How to Test for It<\/strong><\/h4>\n\n\n\n

      Pull the current label from DailyMed and note its revision date. [20] Ask each tool about dosing, contraindications, and boxed warnings. Score each claim as supported, unsupported, or contradicted.<\/p>\n\n\n\n

      What Is a Type 4 Misattribution Error?<\/strong><\/h3>\n\n\n\n

      A Type 4 error assigns a fact to the wrong party. Wolf River’s complaint describes this pattern: an AI answer allegedly attached a lawsuit to the wrong business. [15] In pharma, the equivalent is a trial result or safety warning credited to a competitor’s product in the same class.<\/p>\n\n\n\n

      How to Test for It<\/strong><\/h4>\n\n\n\n

      Ask class-level questions (“which GLP-1 drugs have a boxed warning for thyroid C-cell tumors?”) and check each named product against its own label. [20]<\/p>\n\n\n\n

      What Is a Type 5 Absence Error?<\/strong><\/h3>\n\n\n\n

      A Type 5 error is an omission. In the 5W index, the top 15 companies hold 72.5% of tracked citation share (our sum), so a company outside that group appears in a small share of answers. [8] Absence is not an inaccuracy, but it changes what investors and patients see.<\/p>\n\n\n\n

      How to Test for It<\/strong><\/h4>\n\n\n\n

      Ask neutral category questions without naming your company and record whether your product or company appears.<\/p>\n\n\n\n

      How Do You Run a Pre-Call AI Audit? Methodology<\/strong><\/h2>\n\n\n\n

      This is the protocol we recommend. Another team can reproduce it. It uses public inputs only.<\/p>\n\n\n\n

      Which AI Platforms Should You Test?<\/strong><\/h3>\n\n\n\n

      Test the assistants your audiences use. The 5W index tracked ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews, which is a reasonable starting set. [8] Record the model name and version shown by the tool, the date and time, the country, whether web search was on, and whether you asked for citations.<\/p>\n\n\n\n

      Which Prompts Should You Use?<\/strong><\/h3>\n\n\n\n

      Write three prompt groups. Investor prompts ask about guidance, pipeline timing, and competitors. Clinical prompts ask about indication, dosing, and safety. Neutral category prompts ask about the therapeutic area without naming your company. The 5W method ran each prompt five times per engine in clean sessions. [8] The BMJ team noted that model output is stochastic, which is why a single run proves little. [10]<\/p>\n\n\n\n

      What Source Set Should You Verify Against?<\/strong><\/h3>\n\n\n\n

      Fix the source set before you read any answer. Use the current prescribing information on DailyMed, the latest 8-K and 10-Q on EDGAR, the earnings press release, the call script, and the trial records on ClinicalTrials.gov. [20][21][22] Write down each source’s version date. That step lets a second reviewer reproduce your scoring.<\/p>\n\n\n\n

      How Should You Score Each Answer?<\/strong><\/h3>\n\n\n\n

      Break every answer into single claims. Mark each claim as supported, unsupported, contradicted, or outdated. Then assign a severity.<\/p>\n\n\n\n

      Severity<\/th>Definition<\/th>Action<\/th><\/tr><\/thead>
      S1<\/td>Contradicts a boxed warning, a contraindication, a dose, or a guidance figure.<\/td>Escalate to the owner the same day.<\/td><\/tr>
      S2<\/td>Strips a safe-harbor caveat or misstates a trial result or timeline.<\/td>Escalate to IR and Legal before the call.<\/td><\/tr>
      S3<\/td>Outdated but not contradicted, or credits a competitor.<\/td>Log, and fix the source page.<\/td><\/tr>
      S4<\/td>Absence, vague wording, or missing citation.<\/td>Log for the quarterly review.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      The severity scale is our framework. It is not an FDA or SEC classification.<\/p>\n\n\n\n

      What Is the Audit Timeline Around the Call?<\/strong><\/h3>\n\n\n\n
      When<\/th>Task<\/th><\/tr><\/thead>
      14 days before<\/td>Freeze the prompt library and the source set. Run the first pass.<\/td><\/tr>
      7 days before<\/td>Review S1 and S2 items with the owners. Check that the IR page and label links are current.<\/td><\/tr>
      2 days before<\/td>Run a second pass. Prepare Q&A answers for any recurring error.<\/td><\/tr>
      Call day<\/td>Do not paste any non-public material into any AI tool.<\/td><\/tr>
      1 day after<\/td>Run transcript-summary tests once the replay is posted. Pfizer’s schedule posts the transcript within 24 hours. [17]<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      What Are the Limits of This Method?<\/strong><\/h3>\n\n\n\n

      An audit shows a sample on one day. It does not show what every user sees. Personalization, location, and model updates change answers. The audit does not replace a lawyer’s view on disclosure or a pharmacovigilance review of reportable events.<\/p>\n\n\n\n

      What Does This Mean for Each Function?<\/strong><\/h2>\n\n\n\n

      An audit is useful only if each finding reaches an owner. The table maps functions to the finding types they handle.<\/p>\n\n\n\n

      Function<\/th>Handles<\/th>First question to ask<\/th><\/tr><\/thead>
      Investor Relations<\/td>Types 1 and 2<\/td>Does the AI answer match the latest 8-K and the call preface?<\/td><\/tr>
      Legal<\/td>Types 1 and 2, and any disclosure question<\/td>Could any pre-call input have reached a person named in Reg FD? [3]<\/td><\/tr>
      Regulatory Affairs<\/td>Type 3<\/td>Does the answer match the current label revision?<\/td><\/tr>
      Medical Affairs and Medical Information<\/td>Types 3 and 4<\/td>Does the answer match the label and the published trial?<\/td><\/tr>
      Pharmacovigilance<\/td>Type 3, safety content<\/td>Does any answer understate a labeled risk?<\/td><\/tr>
      Brand and Corporate Communications<\/td>Types 4 and 5<\/td>Is the product named at all in neutral category answers?<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

      What This Means for Investor Relations<\/strong><\/h3>\n\n\n\n

      IR owns the earnings press release and the website that carries it. Q4 Inc. said in March 2026 that it powers IR websites for more than 50% of the S&P 500 and that its AEO for IR Web product adds schema markup to press releases, earnings events, leadership bios, and filings so AI tools can read them. [7] Q4 also reported 30 to 40 point visibility gains in its own early testing, without specifying the metric. [7] Treat that result as a vendor claim until an independent test confirms it.<\/p>\n\n\n\n

      What This Means for Legal<\/strong><\/h3>\n\n\n\n

      Legal should set the rule on what staff may enter into AI tools before an earnings call. Reg FD names the recipients that trigger the rule and lists confidentiality exceptions. [3] Counsel should decide how those provisions apply to each AI vendor contract. EY found that 36% of Fortune 100 companies disclosed AI as a separate 10-K risk factor in 2025, up from 14% in 2024. [16] That change shows legal teams already treat AI as a disclosure topic.<\/p>\n\n\n\n

      What This Means for Medical Affairs, Regulatory Affairs, and Pharmacovigilance<\/strong><\/h3>\n\n\n\n

      These teams own Type 3 and Type 4 errors. The January 14, 2026 FDA and EMA guiding principles address the use of AI to generate evidence across the drug product lifecycle. [18] On our reading, they address sponsor-controlled AI use and do not set rules for consumer chatbot accuracy. A medical-information team should not cite them as a standard for third-party chatbot answers.<\/p>\n\n\n\n

      What This Means for Brand Teams<\/strong><\/h3>\n\n\n\n

      Brand teams own Type 5. Run the neutral category prompts, record who appears, and compare the result with your own share of voice. The 5W index is a template for that comparison. [8]<\/p>\n\n\n\n

      What Should You Do With the Findings?<\/strong><\/h2>\n\n\n\n

      Should You Try to Correct the Chatbot?<\/strong><\/h3>\n\n\n\n

      Start with the source, not the chatbot. Check whether your own pages carry the wrong or stale fact. Update the IR page, the label link, and the press release archive. Then ask the AI vendor to correct a persistent S1 error through its feedback channel. Keep screenshots and timestamps of what you saw. Wolf River’s complaint shows that companies may end up in court over an AI answer, and records of the original output help in that situation. [15]<\/p>\n\n\n\n

      Where Does DrugChatter Fit in This Process?<\/strong><\/h3>\n\n\n\n

      DrugChatter’s AI monitoring (chatmon) re-runs a library of natural-language drug questions across AI models on a recurring schedule and records the answers so teams can compare them with the current label. That covers Type 3 and Type 4 checks at a scale a manual audit cannot reach. It does not replace the primary sources named above. A DrugChatter record shows what a model said. The prescribing information, the filing, and the trial record show what is true.<\/p>\n\n\n\n

      Should You Add Structured Data to Your IR Page?<\/strong><\/h3>\n\n\n\n

      Schema markup helps machines read a page. Q4’s own announcement describes this approach for IR sites. [7] We found no independent study in our research that shows schema markup changes the accuracy of AI answers about a specific issuer, so treat it as a low-cost hygiene step and not a proven fix.<\/p>\n\n\n\n

      Three Scenarios Before the Next Call<\/strong><\/h2>\n\n\n\n

      These scenarios rest on stated assumptions. They are not forecasts.<\/p>\n\n\n\n

      Scenario A: The AI Answer Repeats Last Quarter’s Guidance<\/strong><\/h3>\n\n\n\n

      Assumption: your guidance changed this quarter, and the AI tools index your old release first. The investor who relies on that answer enters the call with the wrong baseline. Action: put the new figure and date in a plain-text sentence near the top of the IR page, and run the Type 1 test the day after the release.<\/p>\n\n\n\n

      Scenario B: The AI Summary of Your Call Drops the Safe-Harbor Preface<\/strong><\/h3>\n\n\n\n

      Assumption: the tool summarizes the transcript for a user. [17] The summary keeps the timeline and drops the qualifier. Action: log it as S2, brief IR and Legal, and consider whether your next call script should state the key qualifier in the same sentence as the forward-looking claim. [5]<\/p>\n\n\n\n

      Scenario C: A Patient-Facing Answer Misstates Your Label<\/strong><\/h3>\n\n\n\n

      Assumption: a model gives a wrong dose or omits a contraindication, as the BMJ and LIU studies found for other chatbots and drugs. [10][11] Action: log it as S1, send it to Regulatory, Medical Affairs, and Pharmacovigilance, and keep the evidence. Pharmacovigilance decides whether any related case report is reportable. The audit does not make that decision.<\/p>\n\n\n\n

      What Are the Limits of This Analysis?<\/strong><\/h2>\n\n\n\n
        \n
      • We did not run a new test of AI answers for this article. Every accuracy figure comes from a cited study with its own model, date, and sample.<\/li>\n\n\n\n
      • Some studies used older models. Results do not transfer to current models without a new test.<\/li>\n\n\n\n
      • The Brunswick figures come from an Axios summary that omits the sample size. [6]<\/li>\n\n\n\n
      • OpenAI’s usage figures come from a secondary summary of OpenAI’s own report. [12]<\/li>\n\n\n\n
      • The 5W index is a PR firm’s estimate with a single quarter of data and an inconsistent publication date. [8]<\/li>\n\n\n\n
      • The Reg FD and safe-harbor sections are our reading of the text. They are not legal advice. Consult counsel.<\/li>\n<\/ul>\n\n\n\n

        Key Takeaways<\/strong><\/h2>\n\n\n\n
          \n
        • Pfizer’s Q3 2026 call was scheduled for November 3, 2026, with a replay and transcript promised within 24 hours. Replays and transcripts are inputs for AI summaries. [17]<\/li>\n\n\n\n
        • 46% of surveyed institutional investors say they are more likely to skip calls and read AI summaries, per Axios’s report of a Brunswick Group study. [6]<\/li>\n\n\n\n
        • In Pew’s March 2025 data, traditional link clicks fell from 15% to 8% of visits when an AI summary appeared. [1]<\/li>\n\n\n\n
        • In the BMJ Quality & Safety study, 26.0% of scored Bing Copilot answers had accuracy issues. [10]<\/li>\n\n\n\n
        • In the LIU study, 10 of 39 ChatGPT answers (25.6%) were satisfactory. [11]<\/li>\n\n\n\n
        • The top 6 companies hold 50.5% of estimated AI citation share in the 5W Q2 2026 index. [8]<\/li>\n\n\n\n
        • Our five-type taxonomy (stale fact, stripped caveat, label drift, misattribution, absence) assigns each AI error to the function that owns the correcting source.<\/li>\n\n\n\n
        • A pre-call audit needs a fixed prompt library, a fixed source set, claim-level scoring, and an escalation path.<\/li>\n<\/ul>\n\n\n\n

          FAQ<\/strong><\/h2>\n\n\n\n

          Does Regulation FD apply to what an AI chatbot says about my company?<\/strong><\/h3>\n\n\n\n

          No provision of Reg FD makes an issuer responsible for a third party’s chatbot output. The rule applies to the issuer’s own disclosures of material nonpublic information to named recipients. [3] Other law may apply to the issuer’s own statements.<\/p>\n\n\n\n

          Can I paste my draft earnings script into ChatGPT to check it?<\/strong><\/h3>\n\n\n\n

          Ask Legal first. A draft may contain material nonpublic information. Reg FD names the recipients that trigger its rule and lists confidentiality exceptions, and an AI vendor’s terms may bear on both. [3] A safer audit uses only public inputs.<\/p>\n\n\n\n

          How often do AI tools get drug information wrong?<\/strong><\/h3>\n\n\n\n

          The answer depends on the model, the date, and the question set. In the BMJ study, 26.0% of 484 scored Bing Copilot answers had accuracy issues. [10] In the LIU study, 29 of 39 ChatGPT answers were unsatisfactory. [11] Neither figure describes current models.<\/p>\n\n\n\n

          Where should I verify a drug claim an AI tool made?<\/strong><\/h3>\n\n\n\n

          Use the current prescribing information on DailyMed and check the revision date. [20] For trial claims, use the ClinicalTrials.gov record and the published paper. [21]<\/p>\n\n\n\n

          What did the Moffatt v. Air Canada decision hold?<\/strong><\/h3>\n\n\n\n

          A British Columbia tribunal held that Air Canada was responsible for information given by a chatbot on its website and rejected the argument that the chatbot was a separate entity. [14]<\/p>\n\n\n\n

          Why does AI citation share matter if my drug is on the label?<\/strong><\/h3>\n\n\n\n

          A label records what is approved. Citation share measures how often an AI engine names a company in answer text. 5W estimates it for 25 companies and finds Lilly at 12.5% and Merck at 6.0%. [8] A company can be absent from answers about its own therapeutic area.<\/p>\n\n\n\n

          Do AI summaries reduce traffic to company websites?<\/strong><\/h3>\n\n\n\n

          Pew found that clicks on traditional links fell from 15% to 8% of visits when a Google AI summary appeared. [1] The study covers general searches by 900 US adults in March 2025, not pharma sites.<\/p>\n\n\n\n

          Does the FDA and EMA AI guidance cover chatbot accuracy?<\/strong><\/h3>\n\n\n\n

          The January 14, 2026 principles aim to inform and promote the use of AI for generating evidence across the drug product lifecycle. [18] On our reading, they do not set accuracy standards for consumer chatbots.<\/p>\n\n\n\n

          What has the SEC done about AI-related statements?<\/strong><\/h3>\n\n\n\n

          On March 18, 2024, the SEC announced settled charges against Delphia (USA) Inc. and Global Predictions Inc., with penalties of $225,000 and $175,000, over statements about their use of AI. [2] Those cases concern claims about a firm’s own AI use.<\/p>\n\n\n\n

          How many people ask AI about health each day?<\/strong><\/h3>\n\n\n\n

          OpenAI reported more than 40 million people submitting health-related prompts to ChatGPT daily, and more than 5% of all ChatGPT messages concerning health. [12] These are OpenAI’s own figures, reported by Medical Economics.<\/p>\n\n\n\n

          References<\/strong><\/h2>\n\n\n\n
            \n
          1. Pew Research Center. (2025, July 22). Google users are less likely to click on links when an AI summary appears in the results. https:\/\/www.pewresearch.org\/short-reads\/2025\/07\/22\/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results\/<\/li>\n\n\n\n
          2. U.S. Securities and Exchange Commission. (2024, March 18). Press release 2024-36: SEC charges two investment advisers over false and misleading statements about their use of artificial intelligence. https:\/\/www.sec.gov\/newsroom\/press-releases\/2024-36<\/li>\n\n\n\n
          3. Legal Information Institute. (n.d.). 17 CFR 243.100: General rule regarding selective disclosure. Cornell Law School. https:\/\/www.law.cornell.edu\/cfr\/text\/17\/243.100<\/li>\n\n\n\n
          4. Legal Information Institute. (n.d.). 17 CFR 243.101: Definitions. Cornell Law School. https:\/\/www.law.cornell.edu\/cfr\/text\/17\/243.101<\/li>\n\n\n\n
          5. Securities Exchange Act of 1934, 15 U.S.C. 78u-5: Application of safe harbor for forward-looking statements. U.S. Government Publishing Office. https:\/\/www.govinfo.gov\/content\/pkg\/USCODE-2022-title15\/html\/USCODE-2022-title15-chap2B-sec78u-5.htm<\/li>\n\n\n\n
          6. Axios. (2026, February 5). How institutional investors are using AI in investment research (reporting on a Brunswick Group study). https:\/\/www.axios.com\/2026\/02\/05\/institutional-investors-ai-investment-research-brunswick<\/li>\n\n\n\n
          7. Q4 Inc. (2026, March 3). Q4 launches AEO for IR Web to help public companies stand out in AI generated answers. Business Wire. https:\/\/www.businesswire.com\/news\/home\/20260303509243\/en\/Q4-Launches-AEO-for-IR-Web-to-Help-Public-Companies-Stand-Out-in-AI-Generated-Answers<\/li>\n\n\n\n
          8. 5W AI Communications. (2026). Pharma and Rx AI Visibility Index 2026. https:\/\/www.5wpr.com\/ai-visibility-index\/pharma-rx-ai-visibility-index-2026\/<\/li>\n\n\n\n
          9. Fierce Pharma. (2026). Eli Lilly, Novo Nordisk top AI citation share as new report questions DTC spend culture. https:\/\/www.fiercepharma.com\/marketing\/eli-lilly-novo-nordisk-top-ai-citation-share-new-report-questions-dtc-spend-culture<\/li>\n\n\n\n
          10. Andrikyan, W., Sametinger, S. M., Kosfeld, F., Jung-Poppe, L., Fromm, M. F., Maas, R., & Nicolaus, H. F. (2025). Artificial intelligence-powered chatbots in search engines: A cross-sectional study on the quality and risks of drug information for patients. BMJ Quality & Safety, 34, 100-109. https:\/\/qualitysafety.bmj.com\/content\/34\/2\/100.full.pdf<\/li>\n\n\n\n
          11. CNBC. (2023, December 5). Free ChatGPT may incorrectly answer drug questions, study says (reporting on a Long Island University study). https:\/\/www.cnbc.com\/2023\/12\/05\/free-chatgpt-may-incorrectly-answer-drug-questions-study-says.html<\/li>\n\n\n\n
          12. Medical Economics. (2026, January 7). 40 million people now use ChatGPT daily for health questions, OpenAI report finds. https:\/\/www.medicaleconomics.com\/view\/40-million-people-now-use-chatgpt-daily-for-health-questions-openai-report-finds<\/li>\n\n\n\n
          13. Euronews. (2026, January 12). Google removes some health-related questions from its AI Overviews following accuracy concerns. https:\/\/www.euronews.com\/2026\/01\/12\/google-removes-some-health-related-questions-from-its-ai-overviews-following-accuracy-conc<\/li>\n\n\n\n
          14. McCarthy Tetrault. (2024, February). Moffatt v. Air Canada: Misrepresentation by AI chatbot (discussing 2024 BCCRT 149). https:\/\/www.mccarthy.ca\/en\/insights\/blogs\/techlex\/moffatt-v-air-canada-misrepresentation-ai-chatbot<\/li>\n\n\n\n
          15. Minnesota Lawyer. (2025, June 13). Solar firm sues Google over AI-generated false claims. https:\/\/minnlawyer.com\/2025\/06\/13\/wolf-river-electric-google-ai-defamation-lawsuit\/<\/li>\n\n\n\n
          16. Niemann, P. (2025, October 28). Cyber and AI oversight disclosures: What companies shared in 2025. Harvard Law School Forum on Corporate Governance (EY). https:\/\/corpgov.law.harvard.edu\/2025\/10\/28\/cyber-and-ai-oversight-disclosures-what-companies-shared-in-2025<\/li>\n\n\n\n
          17. Pfizer Inc. (2026, November 3). Pfizer invites public to join its third-quarter 2026 earnings conference call. https:\/\/s206.q4cdn.com\/795948973\/files\/doc_events\/2026\/Nov\/03\/Pfizer-Invites-Public-Q3-2026-Earnings_FINAL.pdf<\/li>\n\n\n\n
          18. RAPS. (2026, January 14). EMA, FDA issue joint AI guiding principles for drug development. https:\/\/www.raps.org\/resource\/ema-fda-issue-joint-ai-guiding-principles-for-drug.html<\/li>\n\n\n\n
          19. Crowdfund Insider. (2026, April). Survey indicates majority of retail investors using AI to guide decisions (reporting on an Investing.com survey of 938 US investors). https:\/\/www.crowdfundinsider.com\/2026\/04\/271700-survey-indicates-majority-of-retail-investors-using-ai-to-guide-decisions\/<\/li>\n\n\n\n
          20. National Library of Medicine. (n.d.). DailyMed. https:\/\/dailymed.nlm.nih.gov\/<\/li>\n\n\n\n
          21. National Library of Medicine. (n.d.). ClinicalTrials.gov. https:\/\/clinicaltrials.gov\/<\/li>\n\n\n\n
          22. U.S. Securities and Exchange Commission. (n.d.). EDGAR full text search. https:\/\/www.sec.gov\/edgar\/search\/<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"

            Pfizer held its third-quarter 2026 earnings call on Tuesday, November 3, at 10:00 a.m. EST. The company said it would […]<\/p>\n","protected":false},"author":1,"featured_media":1293,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_lmt_disableupdate":"","_lmt_disable":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[1],"tags":[],"class_list":["post-1292","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-general"],"modified_by":"DrugChatter","_links":{"self":[{"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/posts\/1292","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/comments?post=1292"}],"version-history":[{"count":1,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/posts\/1292\/revisions"}],"predecessor-version":[{"id":1294,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/posts\/1292\/revisions\/1294"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/media\/1293"}],"wp:attachment":[{"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/media?parent=1292"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/categories?post=1292"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/drugchatter.com\/insights\/wp-json\/wp\/v2\/tags?post=1292"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}