Source Alignment Checklist for IR Teams
Michael Etheredge is VP of Product Development at QuestionFuel, where he leads how the firm measures and corrects the way AI answer engines describe companies. He was featured by the Southern Economic Development Council on how AI is changing the way companies research communities and regions.
Investor relations teams spend a lot of time shaping the company's story. The website, investor deck, press releases, earnings materials, pipeline updates, FAQs, and leadership messaging all play a role in helping the market understand what matters.
AI has added a new layer. Stakeholders now use AI tools to summarize the company before they ever open the investor deck or visit the website, which means the public record has to be readable by a machine that will compress it into a few sentences.
Key takeaways
- Source alignment is the practice of making every public description of the company consistent, current, and easy for an AI system to interpret.
- AI answers compress many sources into one summary, so inconsistency between the website, the deck, and third-party profiles produces a generic or outdated answer.
- Test the unbranded questions a stakeholder would actually ask, not just your company name.
- Missing context is as damaging as an incorrect fact, because an accurate but incomplete answer still misleads.
- Alignment is a recurring review, not a one-time project, because AI answers shift as sources and models change.
Quick answer: Source alignment is the process of confirming that every public source describing your company says the same current thing, in language clear enough for AI systems to interpret without guessing. When sources disagree, AI tools resolve the conflict on their own, usually by defaulting to whatever is most frequent or most confidently stated rather than what is most current. This checklist walks through the nine reviews an IR team can run to find that risk before the market does.
If public sources are inconsistent, outdated, thin, or hard to interpret, AI systems may produce answers that do not reflect the company's current narrative. This checklist helps IR teams identify where the risk of narrative drift exists.
Is the core company description consistent across sources?
Start with the basics. Does the company describe itself the same way everywhere a stakeholder or a crawler could find it? Review the homepage, About page, investor relations page, corporate overview, investor deck, press release boilerplate, conference profiles, the LinkedIn company page, and third-party database descriptions.
Look for mismatches. If one source describes the company as early-stage, another as clinical-stage, and another by an older business focus, AI has no reliable way to determine which description is current. The company description should be clear, current, and consistent across every one of those surfaces.
Is the lead asset or product explained clearly enough to summarize?
For companies with a lead product, platform, therapy, or solution, the explanation has to survive compression. Ask whether the product is described clearly, whether the mechanism or model is explained in plain language, whether the current stage of development is accurate, whether the target market or patient population is stated, whether the differentiation is easy to identify, and whether outdated descriptions still sit on public pages.
AI tools summarize. If the source material is vague, the summary becomes generic, and a generic summary is functionally the same as being described as an average competitor.
Are recent milestones visible across the public footprint?
Narrative drift often begins when an important update is published but not reflected broadly enough. Review whether recent milestones appear in the places AI systems actually read: trial progress, financing updates, partnerships, regulatory milestones, leadership changes, product launches, market expansion, strategic repositioning, and new data or publications.
A milestone buried in a single press release may not be enough. If an update changes how the company should be understood, it needs to appear across multiple high-value pages.
Which outdated pages are still discoverable?
Old content keeps influencing AI-generated answers long after a team has stopped thinking about it. Search for older public pages, PDFs, archived materials, event descriptions, and third-party profiles that no longer represent the company well: old investor decks, outdated pipeline charts, expired event pages, retired product descriptions, legacy boilerplate, past conference abstracts, old executive bios, and former positioning.
These sources are still visible to search engines and AI systems. They may not need to be deleted, but each one should be reviewed for accuracy and context.
How are competitors being framed in the same answers?
AI answers frequently explain a company in relation to other companies, so IR teams need to understand both how their own company is described and how the category is framed. Ask which companies AI tools name as competitors, whether those comparisons are accurate, whether the company is being left out of relevant category answers, whether a competitor reads as more established because their source material is clearer, and whether the company is being compared to the wrong class, product type, mechanism, or market.
Those answers reveal whether the real problem is visibility, categorization, or narrative clarity. They are three different problems with three different fixes.
Which questions should you actually test?
Do not only test branded queries. Test the questions a stakeholder asks during early research: what does this company do, what should investors know about it, what are the key upcoming milestones, how does it compare to others in the space, what is the market opportunity, what are the biggest risks, what is differentiated about the lead asset or platform, who are the competitors, and which recent updates matter most.
Save the answers and review them against the company's intended public narrative. The gap between the two is the working definition of drift for your company.
Are answers missing context rather than stating wrong facts?
AI answers are not always obviously wrong. Often the larger issue is what they leave out. Look for missing context around timing, differentiation, market need, stage of development, clinical or commercial relevance, recent updates, strategic focus, customer or patient population, and category definition.
A technically accurate answer can still be incomplete in a way that creates a real misunderstanding, and incomplete answers are harder to notice than incorrect ones.
How do you make the strongest sources easier to interpret?
Once gaps are identified, strengthen the sources that carry the most weight. That may mean updating the homepage, improving the investor relations page, adding a plain-language product explainer, creating a market or disease-state overview, improving FAQs, updating press release boilerplate, aligning investor deck language with website language, adding clear milestone context, and creating internal links between related pages.
The goal is not to publish more. The goal is to make the correct narrative easier to find, understand, and repeat.
How often should IR teams re-run this review?
Source alignment is not a one-time project. AI answers change as new sources appear, models update, competitors publish, and search results shift. IR teams should periodically review high-value prompts and track whether the narrative is improving, weakening, or drifting.
The questions worth tracking each cycle: Is AI using current information? Is the company described accurately? Are important milestones included? Are comparisons improving? Are outdated assumptions fading? Is the company appearing in the right category conversations?
Frequently asked questions
What is source alignment in an AI visibility context?
Source alignment is the practice of confirming that every public source describing a company, including the website, investor materials, press releases, and third-party profiles, states the same current facts in language an AI system can interpret without resolving a conflict on its own.
Why does inconsistency between sources matter so much to AI answers?
AI systems compress many sources into a single answer instead of presenting them side by side. When sources disagree, the model picks a version, usually the one that is most frequent, most confidently stated, or most prominent, which is not always the most current.
How often should an IR team run a source alignment review?
A practical cadence is quarterly, plus an additional review before any milestone that changes how the company should be understood, such as a data readout, financing, partnership, or repositioning.
What is the difference between an AI visibility problem and a narrative accuracy problem?
A visibility problem means the company does not appear in relevant answers at all. A narrative accuracy problem means it appears but is described with outdated, incomplete, or miscategorized information. The first is fixed by strengthening signals, the second by aligning sources.
Can outdated pages be left online if they are historically accurate?
Yes, but they should carry clear dates and context so an AI system can tell that they describe a previous stage of the company rather than the current one. Undated legacy material is the most common source of a stale AI answer.
How do you know whether a source alignment fix worked?
Re-run the same set of stakeholder questions after the updates and compare the answers to the earlier set. Improvement shows up as newer facts appearing, older framing fading, and comparisons landing in the correct category.
Final thought
Investor relations has always been about helping the market understand the company. AI does not change that responsibility. It changes where the first impression is formed.
A source alignment review helps IR teams see whether the public record is strong enough for AI systems to interpret the company accurately. When the sources are aligned, current, and clear, the narrative has a better chance of staying intact.
Related: read our main article on AI narrative drift, the DRIFT detection walkthrough, and narrative drift in biopharma during quiet periods.