Should Earnings Releases Use PDF or HTML for AI Search?
An earnings release PDF or HTML for AI search plan is a publishing method for ChatGPT, Google AI Overviews, Google AI Mode and Microsoft Copilot. Publish crawlable HTML beside the approved PDF, with clear financial tables and a direct link to the matching SEC EDGAR filing.
Google can index PDF text, so PDF format alone does not prove why an engine missed or misstated a figure. Before changing the page, preserve the answer, cited sources and exact filing passage. This evidence shows whether the problem involves format, stale information or another source.
Can a PDF earnings release cause AI engines to miss financial figures?
A PDF earnings release cannot be blamed for a missed figure without evidence. The IR team should compare the captured answer, source page and exact filing passage before reaching that conclusion.
Google Search Central lists PDF as a file type that Google can index. PDF format alone does not make an earnings release unavailable to Google Search. Google does not promise that indexing will produce a correct generated answer about every table or figure.
Documented fact: Google can index text from PDF files.
MS House recommendation: Treat format as one item to inspect. Also check whether the engine cited an older release, another page or no source.
Source: Google Search Central, file types Google can index. Last checked: 2026-09-29.
What should the IR team inspect before blaming the PDF?
The IR team should first identify what the engine used. A complete review needs the answer text, source URLs and matching SEC filing passage.
Use this sequence:
- Save the complete answer and test time.
- Record every source URL shown with the answer.
- Identify the financial claim that needs review.
- Find the matching passage in the SEC filing or exhibit.
- Compare the measure, period, unit and basis of presentation.
- Record whether the answer is supported, outdated, unsupported or unclear.
This process separates evidence from a theory about page format. It also helps the IR team find stale sources or inconsistent website copies.
For more review steps, read why AI chatbots get earnings wrong.
How should an IR team handle charts and image-based tables?
An IR team should publish important values as visible text or an HTML table, not only inside an image. This makes each figure easier to identify and verify.
Google advises site owners to provide text alternatives for images. Descriptive alt text can explain an image, but a short description may not contain every value from a financial table.
Keep a chart when it helps readers compare results. Place a labeled table near it when investors need exact values. Identify the measure, period, unit and basis of presentation in visible text.
Source: Google Search Central, image SEO best practices. Last checked: 2026-09-29.
What structured data should an earnings release page use?
Structured data on an earnings release page should describe visible page content. No documented structured data feature guarantees that an engine will interpret every financial figure correctly.
Google documents JSON-LD, microdata and RDFa for structured data on web pages. Google also says structured data must describe content that readers can see on the page.
Documented fact: Google supports structured data on web pages and requires it to match visible content.
MS House recommendation: Add accurate page and publisher data to the HTML page when appropriate. Do not place hidden financial claims in markup.
Source: Google Search Central, understand how structured data works. Last checked: 2026-09-29.
Are format and structured data separate decisions?
Format and structured data are separate decisions for an earnings release page. Crawlable HTML makes the release text, headings, tables and links easier to access.
Valid Article markup can identify page elements such as the headline, publisher and dates. It cannot guarantee that ChatGPT, Google AI Overviews, Google AI Mode or Microsoft Copilot will cite the release.
MS House recommends fixing access first. Then add markup that matches the visible content. Read earnings release structured data and AI search for the markup review.
How should an earnings press release be formatted for AI engines?
An earnings press release should use crawlable HTML beside the approved PDF. The HTML should include visible text, descriptive headings, labeled tables and links to the PDF and SEC filing.
MS House does not claim that this format guarantees a correct answer. It gives readers and crawlers a clear official page to inspect. Both versions should present the same approved information without changing its meaning.
Should an earnings release also appear as crawlable HTML?
Yes. A crawlable HTML earnings release should appear beside the approved PDF. The main release text and financial tables should not require a PDF viewer.
Google explains that page content should be available in rendered HTML for indexing. HTML also gives headings, paragraphs, tables and links defined roles on the page.
Documented fact: Google needs access to page content and crawlable links to discover and understand web pages.
MS House recommendation: Put the release text and financial tables in HTML. Keep the approved PDF available as a separate company copy.
Sources: Google Search Central, JavaScript SEO basics and Google Search Central, crawlable links. Last checked: 2026-09-29.
Does the HTML version replace the PDF?
No. The HTML version should complement the approved PDF and SEC submission. It should not replace the company’s legal, accounting or disclosure process.
The PDF can remain available for downloads and company archives. The HTML page gives investors and search crawlers another official way to read the release.
MS House recommends giving each release its own permanent HTML URL. Do not overwrite an older release with new results. Show the release date and covered reporting period as visible text.
What should the HTML earnings release page contain?
The HTML earnings release page should contain the approved title, release date, narrative and financial tables. It should also link to the PDF and matching SEC record.
Use headings that name each financial topic. Avoid headings that provide no context when read alone. Each table should identify its measures, periods and units.
MS House also recommends a self-referencing canonical link on the HTML page. Google documents canonical URLs as a way to identify a preferred page among duplicate or similar pages.
Source: Google Search Central, canonical URL guidance. Last checked: 2026-09-29.
How should an IR team format GAAP and adjusted figures?
An IR team should label GAAP and adjusted figures separately. Each figure should include an exact measure name, period, unit and basis of presentation.
Avoid labels that require a reader to infer whether a figure is GAAP, non-GAAP, reported or adjusted. Use the approved terms consistently across the HTML page, PDF and submitted exhibit.
For how engines report these measures, read how AI engines handle GAAP and adjusted EPS.
The SEC publishes rules and staff guidance about non-GAAP financial measures. The company’s legal and accounting teams must decide how those requirements apply to its release.
Documented fact: The SEC provides requirements and guidance for presenting non-GAAP financial measures.
MS House recommendation: Use clear labels and keep each adjustment close to its explanation. MS House does not provide legal or accounting advice.
Source: SEC, Non-GAAP Financial Measures. Last checked: 2026-09-29.
Should financial figures use HTML table elements?
Yes. Financial figures organized into rows and columns should use real HTML table elements. Header cells should identify each measure and period.
A useful table includes:
- A descriptive caption
- Labeled rows and columns
- Visible units
- Clear period names
- Separate GAAP and adjusted labels
- Footnotes placed near the figures they explain
If a table reports amounts in millions, state that near the table. If a value is per share, label it as a per-share value. Do not make readers infer the unit from nearby prose.
For table-level guidance, read how to format earnings tables for AI search.
How should reconciliations appear in HTML?
A reconciliation should appear in visible HTML when it is part of the approved release. The HTML version should preserve the approved labels, order and figures.
MS House recommends placing the reconciliation near the adjusted measure it explains. Keep footnotes next to the relevant table or immediately below it.
The SEC filing or exhibit remains the source used to verify the web copy. If the HTML page differs from that record, the IR team should review the difference through its approved process.
How should an IR team format financial tables so AI engines can read them?
An IR team should format earnings release financial tables as semantic HTML. Each table should have clear headers, visible units and separate GAAP and adjusted labels.
Use <table>, <caption>, <th> and <td> elements. Label every row and column with its measure, period and unit. Do not rely on color or an image to explain a figure.
Add a short text summary near the table for readers, screen readers and crawlers. Keep footnotes beside the figures they explain.
Link the table to the matching SEC filing or filed exhibit with a crawlable <a> link. The SEC filing remains the record used to verify the figures.
This format does not guarantee that an AI engine will use, interpret or cite the page.
Sources: SEC EDGAR search, SEC, Structured Data and Google Search Central, crawlable links. Last checked: 2026-09-29.
What should remain in the filed exhibit?
The filed or furnished exhibit should retain the complete disclosure approved through the company’s reporting process. A website copy should not replace the SEC submission or change its meaning.
Form 8-K Item 2.02 addresses disclosure of results of operations and financial condition. The company must confirm whether its release is filed or furnished, along with the correct exhibit designation.
MS House uses SEC EDGAR filings as financial ground truth. The relevant source may be an 8-K, 10-Q, 10-K or an exhibit. The correct record depends on the claim under review.
Sources: SEC Form 8-K and SEC EDGAR search. Last checked: 2026-09-29.
How should the website connect to the filing?
The website should link the HTML release to its matching SEC filing or exhibit. The link should use a standard HTML anchor that a crawler can follow.
Check the company name, filing date and accession before publishing the link. Label the link by destination, such as “View the SEC filing” or “View the submitted exhibit.”
Google documents that crawlable links generally use an <a> element with an href attribute. SEC EDGAR provides the public filing record.
Sources: Google Search Central, crawlable links and SEC EDGAR search. Last checked: 2026-09-29.
| Artifact | Main role | Verification source |
|---|---|---|
| SEC filing or exhibit | Public SEC record | SEC EDGAR |
| IR website HTML page | Readable web copy | Matching filing or exhibit |
| PDF release | Downloadable company copy | Matching filing or exhibit |
Can the HTML page include extra commentary?
An HTML release should not add financial claims that have not passed the company’s approval process. Any added explanation should follow the company’s disclosure controls.
Navigation labels, accessibility text and descriptive links can help readers use the page. They should not alter the financial message.
The company’s legal, accounting and disclosure teams should decide what may be added. MS House checks public answers against public filings and does not provide legal advice.
How can an IR team help search crawlers find the latest release?
An IR team should give each release a permanent URL and link it from the IR website. MS House also recommends adding indexable release pages to the XML sitemap.
Google documents sitemaps as a way to provide information about pages and files. Google also states that a sitemap does not guarantee crawling or indexing.
The page should return a successful HTTP response and remain available after a newer release appears. Show the release date in visible text and use a self-referencing canonical link.
Source: Google Search Central, sitemaps overview. Last checked: 2026-09-29.
Should the IR site allow OAI-SearchBot?
Yes, if the company wants its public earnings pages to be eligible for ChatGPT search. OpenAI identifies OAI-SearchBot as the crawler used to surface websites in search results.
OpenAI documents OAI-SearchBot separately from GPTBot and provides robots.txt controls for each crawler. The IR team should confirm that robots.txt allows OAI-SearchBot to reach the directory that contains public releases.
The crawler review should cover the permanent HTML release URL, not only the downloadable PDF. For step-by-step robots.txt and OAI-SearchBot checks, read the OAI-SearchBot guide for investor relations.
Source: OpenAI crawler documentation. Last checked: 2026-09-29.
Do Google AI features need special optimization?
No. Google states that no special AI file or markup is required for Google AI Overviews or Google AI Mode.
Google advises site owners to follow its standard search guidance. Pages must meet the normal technical requirements for Google Search. Structured data should also match visible page content.
MS House recommends accessible pages, accurate text, descriptive headings and crawlable links. Do not add hidden text or unsupported financial claims for search systems.
Source: Google Search Central, AI features and your website. Last checked: 2026-09-29.
Which AI engine gives the most accurate public company financial answers?
This page does not contain evidence that ranks ChatGPT, Google AI Overviews, Google AI Mode or Microsoft Copilot for financial accuracy. An IR team should test the same company-specific topics on each engine and verify every claim against the matching SEC filing.
Preserve each answer, its source URLs, the test time and the exact filing passage. Apply the same evidence standard to every engine because answers and cited sources can change.
Is there data on how often AI engines get earnings figures wrong?
This page does not provide a measured error rate or rank the most common error types. It contains no research sample that could support either claim.
A study of error frequency would need a stated collection date, sample size, engine list and verification method. Each finding would also need the captured answer, cited source and matching SEC filing passage. Without that evidence, the IR team should not present a general error rate to the board.
Where does Google AI Overviews get quarterly revenue and EPS figures?
An IR team should inspect the source links shown with each Google AI Overviews answer. Do not assume that every revenue or earnings-per-share figure came from the latest SEC filing or IR page.
Save the full answer and every cited URL. Then compare each figure with the matching SEC filing passage. Record whether the answer is supported, outdated, unsupported or unclear.
The source review matters more than a general claim about where Google gets financial figures. The evidence should show which page supported the answer at the recorded test time.
How should the team test investor answers after publishing?
The IR team should test ChatGPT, Google AI Overviews, Google AI Mode and Microsoft Copilot. Each test record should include the engine, time, answer, source URLs and matching filing text.
Test topics can cover reported results, GAAP measures, adjusted measures, guidance, segments and leadership facts. MS House does not use one fixed public question list. A prediction model creates a company-specific set and updates it each quarter.
Do not treat an answer as verified because it sounds reasonable. Compare each financial claim with the relevant SEC filing passage.
When should an IR team test investor answers?
An IR team should test investor answers after publication and again after crawlers have had time to revisit the site. No official source promises a fixed time for an updated page to appear in generated answers.
Record the time of each test because answers and cited sources can change. Preserve the original answer before writing a summary or assigning a finding.
The investor relations AI search checklist provides a related publishing and evidence process.
What evidence should each test preserve?
Each test should preserve the answer text, engine name, test time, source URLs and exact filing passage used for verification. The evidence should let another reviewer repeat the comparison.
Use clear finding labels:
- Supported: The filing supports the answer.
- Outdated: The answer uses information that a newer filing replaced.
- Unsupported: The available filing does not support the claim.
- Unclear: The evidence does not support a firm conclusion.
For the full process, read how MS House checks investor answers against SEC filings.
What should the team document for the CFO?
The CFO summary should begin with the most important mismatch, its source and the exact filing passage. Keep the action list short and assign an owner to each proposed fix.
A useful record separates evidence from recommendations. Evidence includes the answer, cited URL and filing passage. Recommendations identify the page, responsible team and proposed correction.
Do not claim that an answer caused a market, legal or disclosure effect without a reliable public source. Report only what the evidence supports.
What should the evidence table include?
The evidence table should include enough detail for finance, legal, IR and communications teams to repeat the check.
| Field | What to record |
|---|---|
| Engine | ChatGPT, Google AI Overviews, Google AI Mode or Microsoft Copilot |
| Test record | Test time and complete captured answer |
| Source | Every cited URL shown with the answer |
| Filing evidence | Form, filing date, exhibit and exact passage |
| Finding | Supported, outdated, unsupported or unclear |
| Action | Page to review and responsible team |
A score without the answer, source and filing passage is not enough for review. The CFO should see what was said, what the filing says and what the team recommends next.
Frequently asked questions
Should an earnings release be HTML or PDF?
MS House recommends publishing both. Keep the approved PDF available, then publish a matching HTML version with visible text, labeled tables and a direct link to the SEC filing.
Does a PDF stop Google from indexing an earnings release?
No. Google Search Central lists PDF as an indexable file type. Indexing does not prove that every table, chart or relationship between figures will be interpreted correctly.
Do Google AI features require special structured data?
No. Google Search Central says no special markup is required for Google AI Overviews or Google AI Mode. Any structured data should describe content that readers can also see.
When should an IR team test investor answers?
MS House recommends testing after publication and again after crawlers have had time to revisit the site. Record the engine, answer, source URLs, test time and matching filing text.
See the fictional sample audit to review how MS House presents answers, filing evidence and proposed fixes.