Picture a portfolio manager with a call on her calendar at 2 p.m. Your CEO is on the other end. At 1:45 she opens an AI assistant and types, "What's going on at this company?" Thirty seconds later she has a clean, confident summary.
Where did that summary come from? Probably a mix of your filings and a few news stories, plus whatever analyst commentary the model could reach. Your latest investor presentation might be in there, or it might be missing. Your own words might be buried under someone else's take.
This is already happening across the IR community. It's a good question to look into before your next earnings cycle, because the answer tells you how much of your story investors are getting without you in the room.
Your IR website was built for people who click around. Investors still do that, but now, many of them also ask an AI tool for a first read, then use your site to check it.
That changes the job a little. Your site and your releases feed the summary before anyone reads them directly. If those sources are stale or hard to parse, the summary will reflect that.
The risk is simple. An investor walks into a meeting with a version of your company that is six months old, or that leans on a skeptical analyst note. You spend the first ten minutes correcting the record when you could be building the relationship.
For another take on how investors use AI tools to research companies, this piece is a good read.
Before you change anything, find out where you stand. This will take about an hour, but it will be a useful hour you'll spend this month.
Pick three AI assistants that your investors are likely to use. Ask each one the questions an investor would ask.
Save the answers. Then compare them to your most recent disclosure, line by line. You're looking for facts that are out of date and framing that came from a third party.
Most teams find a few surprises. A product line you retired two years ago could show up as a growth driver or a strategic priority you lead every call with doesn't appear at all.
An AI tool works with whatever it can reach and understand. Your IR site was built for people, so a few behind-the-scenes fixes can help. You don't need to write any code. These are questions to take to your web team.
Schema markup. Think of it as a label on the back of each page. You see a headline, a date and a table. A crawler sees a block of text and has to guess what it all means. Schema markup adds a short note in the page's code that says, "This is an earnings call, it happens on this date and here is the webcast link." The standard comes from a site called schema.org, and the note is usually written in a format called JSON-LD, which your web team will recognize. A few labels fit IR well.
Robots.txt. This is a small text file on your website that tells automated visitors, such as search engines and AI crawlers, which pages they can read and which to skip. Most IR teams have never looked at theirs, and IT often set it up years ago. If it blocks the crawlers behind AI tools, those tools may skip your site and lean on third-party sources. Ask your web team to show you what's in it. Which crawlers to allow is a call for IT and legal to make together, and it's better made on purpose than by default.
The basics. Publish presentations and transcripts as web pages next to the PDFs, since PDFs are harder for crawlers to read well. Keep your sitemap current with accurate dates. Make sure your company name, ticker and executive names match across data aggregators, your SEC filings and reference pages, because small mismatches can split one company into two.
An AI tool works with whatever it can reach and parse. Most IR sites were built for people, so a few technical fixes go a long way.
You can start with structured data. Schema.org markup, usually added as JSON-LD, tells crawlers exactly what a page contains. A few types are especially useful for IR.
Next, check the basics.
Publish presentations and transcripts as HTML pages alongside the PDFs, because PDFs are harder for crawlers to read well.
Keep your sitemap current with accurate last-modified dates.
Ask your web team whether your robots.txt rules allow the AI crawlers you want reading your site. That's a decision to make with IT and legal, and it's better made on purpose than by default.
Then look beyond your own domain. Models lean heavily on data aggregators, your SEC filings and reference pages, so make sure your company name, ticker and executive names match everywhere. Small mismatches can split one company into two entities.
Let's be honest about the limits. No IR team gets to dictate what a third-party model says. Models pull from many places, and you won't win every disagreement with a news article or a research note.
What you can do is make your own material the clearest and most current source on the topic. That shifts the odds. When the facts you published are easy to find and consistent across channels, summaries tend to follow them.
It also helps to know when a model gets something wrong. If you catch a factual error in an AI answer, check where it came from. Often the source is an old page or a stale data feed that you can correct or clarify. Fixing the source is more durable than arguing with the output.
Treat this like any other part of your earnings routine. A simple rhythm works well.
That last step pays off beyond IR. When your CEO knows how the company is being described, they can prepare for the question behind the question. If the model keeps calling your margins "volatile," you know to address that before someone raises it.
You don't need a big project to get started. Run the question set this week and see what comes back. Write your equity story in plain words. Check that your most current materials are easy to find.
Investors have always formed a view of your company before they talk to you. The difference now is that the first draft of that view is written in seconds, by a tool you didn't choose. The teams that pay attention to it early will spend less time correcting the record and more time telling their story.
If you want to go deeper on getting your story right in AI search, check out Q4's AEO playbook. It walks through the practical steps in more detail.