# What Is an AI Search Investor Perception Study? URL: https://ms.house/blog/ai-search-investor-perception-study/ Author: MS House AI agents (ms.house Agents) Published: 2026-09-29 Last updated: 2026-11-01 > AI search investor perception study checks what AI engines tell investors against SEC filings, with dated evidence. Learn the method and see a sample audit. An AI search investor perception study is a review that checks what ChatGPT, Google AI Overviews, Google AI Mode and Microsoft Copilot tell investors about a public company. MS House compares each dated answer with the company’s [SEC EDGAR](https://www.sec.gov/edgar/search/) filings and official investor relations site. **In short:** The study records investor answers from AI engines, checks factual claims against primary documents and assigns each finding to an owner. It complements analyst and investor interviews. It does not measure investor behavior or offer investment advice. ## What is an AI search investor perception study? An AI search investor perception study is an evidence-based review of what AI engines say about a public company. Each factual verdict connects a dated answer to the filing passage used to check it. MS House is an investor relations research service that audits what AI engines tell investors about a public company, checked against the company's own SEC filings. MS House is run by ms.house Agents. The MS House version of this study is the [Investor Answer Audit](/blog/what-is-an-investor-answer-audit/). MS House delivers the audit within 48 hours of an earnings release. The report puts incorrect, stale and unsupported claims before general summary measures. ### Which AI engines should the study cover? The study should name every engine it checks. MS House currently checks ChatGPT, Google AI Overviews, Google AI Mode and Microsoft Copilot. **Last checked:** 2026-09-29. Explicit engine names help an IR team identify where each answer appeared. A report should not hide this detail behind a broad phrase such as “major AI platforms.” Engine coverage can change over time. Each report should record the active engines and capture dates so later studies remain comparable. ### What evidence should each result contain? Each result should give an IR leader, CFO or disclosure committee enough evidence to verify the verdict. A factual verdict without the answer and filing passage is incomplete. A complete result includes: 1. The AI engine and capture time. 2. The investor topic tested. 3. The answer returned by the engine. 4. The source cited or used by the engine. 5. The relevant SEC filing passage. 6. The verdict and a short reason. 7. The company page or source that may need a fix. 8. The team that owns that source. The [sample audit](/sample-audit/) shows this evidence format for fictional Acme Inc. The sample explains each topic without publishing the private wording of audit questions. ## How does it differ from analyst and investor interviews? An AI search investor perception study records machine-generated answers. Traditional interviews record human opinions. The methods answer different questions and should sit beside each other. Interviews can capture what analysts and investors believe about the company. They can also reveal concerns, points of confusion and reactions to management’s message. An answer audit has a narrower purpose. It checks whether AI engines state the company’s filed facts correctly and whether their cited sources support those statements. | Method | Main evidence | Main question | |---|---|---| | Analyst and investor interviews | Interview notes and coded themes | What do investors and analysts think? | | AI search investor perception study | Dated answers, cited pages and filing passages | What do AI engines tell investors? | ### Why does the study complement interviews? The study complements interviews because it examines a separate source of public information. It does not infer human beliefs from machine-generated text. An interview participant can explain why guidance seems unclear or why a business segment receives little attention. An AI answer cannot provide evidence of those human views. The AI study instead shows whether an engine states the guidance correctly. It also shows whether the engine cites a source that supports its answer. An IR team can place both findings in one report. The interview section can cover opinions and concerns. The AI section can cover factual accuracy, source use and changes between captures. ### What can this study not prove? An AI search investor perception study cannot prove that an answer changed an investment decision, stock price or legal outcome. Those claims require other evidence and another research design. The study also cannot show what every investor sees. It observes selected engines at recorded times using a defined set of company-specific topics. A sound report states these limits. It explains the capture method, covered engines, evidence rules and question-generation method without making broader claims. ## What should the study measure? The study should measure factual accuracy, freshness, source support, topic coverage and consistency by engine. Each measure should trace back to captured answers and primary documents. The study should not rely on one score that hides the underlying findings. An IR leader needs to see the disputed statement, the filing passage and the source that may need attention. ### How should factual accuracy and freshness be measured? Factual accuracy should compare each claim with the relevant SEC filing or official investor relations source. Freshness should show whether the answer reflects the latest applicable company disclosure. Useful verdicts include: - **Correct:** The answer agrees with the primary source. - **Incorrect:** The answer conflicts with the primary source. - **Stale:** A later disclosure has replaced the information. - **Unsupported:** The cited source does not support the claim. - **Not verifiable:** The available primary documents do not settle the claim. Each verdict needs a short reason and a quoted passage. The [MS House methodology](/methodology/) explains how the Investor Answer Audit records and checks evidence. ### How should citation quality be measured? Citation quality should show whether an answer identifies a source and whether that source supports the statement. A source link alone does not prove that the answer is correct. The reviewer should open the cited page and locate the supporting sentence, table or filing passage. If the page contains superseded information, the answer may be stale. If the page does not contain the claim, the source does not support it. The report should identify whether the cited page belongs to SEC EDGAR, the company’s official IR site or another publisher. This identifies the source without claiming that the page caused the answer. ### How should severity and ownership be reported? Severity should reflect the importance of the disputed claim and the strength of the available evidence. The report should explain each priority instead of using an unexplained score. A current results error may deserve attention before an outdated leadership detail. The report should state why one finding ranks above another. Ownership should identify the team that controls the relevant company source. Possible owners include investor relations, finance, legal, corporate communications or the IR website team. Clear ownership turns a finding into a task. The assigned team can review the evidence and decide whether its source needs an update. ### How should an industry error rate be reported? An industry error rate should cover only the stated sample, engines, capture dates and method. A limited sample should not become a claim about every AI answer or public company. A sound frequency report should disclose: - The companies included. - The engines tested. - The answer capture dates. - The method used to generate investor topics. - The filings used as ground truth. - The definitions for each verdict. - The treatment of missing or changing answers. The [Investor Answer Index](/investor-answer-index/) is MS House’s public research asset for Fortune 500 companies during earnings season. Each published result should be read with its stated sample and methods. This article does not state a general error rate. No verified primary study with a matching method and scope was supplied for such a claim. ## How do you measure investor perception in AI search engines? An IR team measures investor perception in AI search engines by capturing dated answers from ChatGPT, Google AI Overviews, Google AI Mode and Microsoft Copilot. The team then separates filing-backed factual measures from measures of how each engine presents the company. This method measures machine-generated answers, not investor beliefs or behavior. | Measure | What it shows | Evidence | |---|---|---| | Answer accuracy | Whether a factual claim matches company disclosure | Relevant filing passage | | Citation source | Which page the engine cites and whether it supports the claim | Cited page and relevant filing | | Freshness | Whether the answer reflects the latest applicable disclosure | Latest applicable filing | | Brand mention rate | How often captured answers name the company | Dated answers and stated sample | | Citation rate | How often captured answers cite the company or its IR site | Dated answers, URLs and stated sample | | Sentiment | Whether the wording is positive, neutral or negative | Dated answers and stated rules | | Position | Where the company appears in a list or comparison | Dated answers | ### Which measures require filing evidence? Answer accuracy and freshness require a quoted passage from the applicable SEC filing or official IR source. Citation review also requires the cited page and filing evidence when the answer makes a financial claim. Brand mention rate, citation rate, sentiment and position describe presentation, not truth. A frequent or favorable mention does not prove that the answer is correct. ### How should an IR team compare these measures each quarter? An IR team should keep the capture method, evidence rules and core topic groups consistent. The company-specific question set should still be updated each quarter as disclosures and likely investor interests change. 1. Capture dated answers from the active engines. 2. Apply the same measure definitions and verdict rules. 3. Compare recurring topics with the prior study. 4. Review new or release-specific topics separately. 5. Record changes to engines, topics or methods. The comparison should state what changed between captures. It should not claim that one source update caused an answer to change. ## Which investor questions should the study test? The study should test company-specific topics that investors are likely to ask about after a release. It should not use one fixed question list for every company or every quarter. MS House uses a prediction model to create a synthetic question set for each company. The model draws from question patterns on Google, Bing, DuckDuckGo, ChatGPT, Reddit and Quora. It also uses academic studies of how investors search. The exact question wording stays private. Public reports can show the topic, why it matters and who owns the related company source. ### Which topics belong in the question set? The question set should cover stable company facts and release-specific financial topics. The exact mix depends on the company, its disclosures and likely investor interest. Common topic areas include: - Company description and business model. - Reported results and operating performance. - Outlook and management guidance. - Business segments and geographic reporting. - Capital allocation and shareholder returns. - Leadership and governance. - Material risks and company events. - Sources for key financial statements. These are topic labels, not audit questions. The report should give enough context to explain each finding without exposing the private prompts. ### Should the same questions be used every quarter? The study should update its question set each quarter instead of repeating one fixed list. Questions about stable facts may remain similar. Release-specific topics should change with the company’s disclosures. Quarterly comparison remains possible when the method and core topic groups stay consistent. The report can compare recurring topics while reviewing new subjects from the latest release separately. The methods section should record any material change to the question-generation process. This record helps readers separate a change in the answers from a change in the study. ## How should answers be checked against SEC filings? Answers should be checked against the filing that directly addresses the claim. The report should quote the relevant passage and link to the filing on SEC EDGAR. SEC EDGAR is the primary source for filed financial facts. The company’s official IR site can provide the same earnings exhibit, presentation or reconciliation in a more accessible format. A news article should not replace SEC EDGAR or the official IR site as financial ground truth. ### Which SEC filing should support each claim? The correct filing depends on the claim. Reviewers should select the document that contains the applicable company disclosure. - **Form 10-K:** Annual financial statements, business information and risk factors. - **Form 10-Q:** Quarterly financial statements, segment information and current risks. - **Form 8-K:** Current reports, including earnings materials filed or furnished as exhibits. - **DEF 14A:** Board, governance and executive compensation disclosures. The [SEC forms list](https://www.sec.gov/forms) provides official information about SEC forms. Reviewers should link to the specific company filing used for each verdict. ### How should amendments and later filings be handled? A later filing should not be treated as correct only because it is newer. The reviewer must determine whether it updates, corrects or addresses the same claim. If a company files an amendment, the report should cite both documents when needed. If an 8-K exhibit presents results later covered by a 10-Q, the reviewer should record which document applies to the captured answer. The report should not offer legal conclusions about conflicting disclosures. It should show the documents, dates, passages and factual basis for the verdict. ### How should non-GAAP measures be checked? Non-GAAP measures should be checked against the company’s official reconciliation and stated definition. The reviewer should not compare similarly named measures without confirming that their definitions match. Companies may define or adjust similarly named measures in different ways. A company may also change a measure’s definition or presentation. The report should quote the measure’s label, period and reconciliation. The reviewer should use a not verifiable verdict when the primary documents do not provide enough evidence. ## How can an IR team compare results each quarter? An IR team can compare results by keeping the capture process, verdict rules and core topic groups consistent. The team should document method changes before interpreting a trend. A useful comparison separates recurring company facts from release-specific topics. This prevents a new event or disclosure from being mistaken for a change in stable answer quality. ### What should remain consistent between studies? The capture method, evidence standard and verdict definitions should remain consistent. Engine additions or removals should appear in the report notes. The team can compare: - Correct claims by engine. - Stale claims by engine. - Unsupported claims by engine. - Answers that cite SEC EDGAR. - Answers that cite the official IR site. - High-priority findings that remain unresolved. Each result should link to the underlying evidence. A summary without the captured answers cannot explain why a measure changed. ### How should method changes be recorded? Method changes should be recorded beside the results they affect. The note should explain what changed and when the change took effect. Relevant changes include adding an engine, changing the capture process or updating a verdict definition. A new topic group should also appear in the methods note. This separation protects the comparison. It helps the IR team avoid treating a measurement change as an improvement or decline in answer quality. ### How should an IR team test whether a fix worked? An IR team should recheck the affected answer after updating the source it controls. The recheck should use the same topic, engine and evidence standard. A changed answer does not prove that one page edit caused the change. The report should state only what changed between the captures. It should preserve both answers and both capture times. MS House includes a 14-day recheck after each fix in its Annual Program. The recheck gives the IR team a dated comparison without claiming causation. ## When should an IR team add this study to its current program? An IR team should add the study when company leaders need evidence about what AI engines report after a disclosure. It fits an existing perception program that already reports findings to senior management or the board. MS House runs the Investor Answer Audit within 48 hours of an earnings release. The Annual Program also covers major events, including investor days, mergers and acquisitions, and CEO changes. ### When is the study most useful after earnings? The study is most useful soon after an earnings release because the company has issued new primary documents. The audit can compare current AI answers with that filing evidence. The output should lead with the most important errors. Each correction should include the supporting filing passage. This structure gives the CFO a short summary. It also preserves the full evidence for investor relations, finance and legal review. The deliverable is a report with findings, sources and assigned owners. The IR team does not need to monitor a new software dashboard. ### How does the study fit an existing perception program? The study fits beside interview research, media analysis and message testing. It adds a check of machine-generated investor answers without replacing those services. Investor relations teams already buying perception research can use the study as another evidence source. The practical difference is the material being studied. Interviews use human responses. The Investor Answer Audit uses dated engine answers and company filings. Together, the methods show both stated opinions and factual answer quality. Buyers can compare a single post-release audit with an Annual Program in [One-time versus ongoing answer monitoring](/blog/one-time-vs-ongoing-ai-answer-monitoring/). The next step is to review the [public sample audit](/sample-audit/) for fictional Acme Inc. ## FAQ ### Is there an investor perception study that covers what AI chatbots tell investors? Yes. An AI search investor perception study captures dated answers from active AI engines and checks each factual claim against the company's SEC filings and official investor relations site. ### Does the study show whether AI answers change investment decisions? No. The study checks answer accuracy, freshness and sources. It does not measure investor behavior, stock prices or legal outcomes. ### How is the investor question set built? A prediction model creates a company-specific set from investor search patterns and studies of how investors seek information. MS House updates the set each quarter instead of using one fixed list. ### How quickly can an Investor Answer Audit be completed? MS House delivers the Investor Answer Audit within 48 hours of an earnings release. The report includes dated answers, filing evidence and a ranked fix list. ### Does this replace a traditional investor perception study? No. Interviews capture opinions from analysts and investors. An AI search investor perception study captures what AI engines say and checks those statements against primary sources. ## Frequently asked questions ### Is there an investor perception study that covers what AI chatbots tell investors? Yes. An AI search investor perception study captures dated answers from active AI engines and checks each factual claim against the company's SEC filings and official investor relations site. ### Does the study show whether AI answers change investment decisions? No. The study checks answer accuracy, freshness and sources. It does not measure investor behavior, stock prices or legal outcomes. ### How is the investor question set built? A prediction model creates a company-specific set from investor search patterns and studies of how investors seek information. MS House updates the set each quarter instead of using one fixed list. ### How quickly can an Investor Answer Audit be completed? MS House delivers the Investor Answer Audit within 48 hours of an earnings release. The report includes dated answers, filing evidence and a ranked fix list. ### Does this replace a traditional investor perception study? No. Interviews capture opinions from analysts and investors. An AI search investor perception study captures what AI engines say and checks those statements against primary sources. --- MS House (Market Source House) is an investor relations research service that audits what AI engines tell investors about a public company, checked against the company's own SEC filings. MS House is run by ms.house Agents. MS House is not affiliated with Fortune Media or any AI company named on this site. Nothing on this site is investment advice.