Your brand shows up in AI answers. Now what?

Maybe a competitor gets recommended more often. Maybe the answer describes a pricing plan you dropped last year. Maybe AI bots can't even read your best pages. Each problem needs a different fix, and guessing wrong costs you months.

Look for patterns, not single answers

AI answers change from one run to the next, so one good result proves very little.

A better approach is to track three kinds of questions. The first covers problems your customers want solved, which shows how people discover brands. The second covers comparisons, which show who AI recommends and to whom. The third covers facts about your business, which tells you whether AI gets your pricing and features right. Over time, the same sources and the same errors will keep appearing, and those are what you investigate.

Be careful with the numbers, though. Share of voice scores from tracking tools are modeled estimates. They shift depending on which competitors you include, so add one to the list and your share can shrink without anything changing. Use them to spot trends, not as a market share figure.

Check that bots can read your site

Before you rewrite anything, make sure AI crawlers can reach your pages. Look at firewall rules, broken links, timeouts and pages that depend on JavaScript.

The JavaScript point is well supported. A Vercel and MERJ study found that none of the major AI crawlers it tested render JavaScript, with Gemini the exception because it uses Google's crawling infrastructure. If your key content only appears after scripts run, most AI bots may see a blank page. That study is from December 2024, so test your own site instead of assuming.

Robots.txt is the next thing to review, because OpenAI runs separate bots for separate jobs. OAI-SearchBot powers ChatGPT search, and if you block it you won't appear in those answers. GPTBot is used for training, so you can block it without affecting search. ChatGPT-User acts on behalf of a person, and robots.txt rules may not apply to it at all.

Your CDN deserves a look too. Reports say Cloudflare introduced new default settings on September 15, 2026, blocking training and agent crawlers on pages with ads while leaving search crawlers allowed. This reportedly applies to new domains and free-plan customers. If you're on Cloudflare, check your settings, because a friendly robots.txt won't help if the request is stopped before it reaches your site.

Fix your own information first

Start with what you control: pricing pages, product descriptions, profiles and old blog posts. Outdated or conflicting details are a common reason AI describes your product wrongly, and they're the cheapest thing to fix.

Mistakes on other websites are harder. You can contact the authors of pages that get cited often, and it's worth doing for the ones you can realistically reach. Expect it to be slow and for many requests to go unanswered. If a correction isn't possible, publish a better and more current page yourself. And if the same error shows up in many places, one new article probably won't be enough to override it.

Where JSON-LD fits in

JSON-LD is a way of labeling the facts on a page so machines don't have to guess. It sits in a script block in your HTML and uses the Schema.org vocabulary. For the problem of AI getting your pricing wrong, it's a natural thing to reach for.

Here is a simple example for a fictional software product:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "Northwind Analytics",
  "applicationCategory": "BusinessApplication",
  "operatingSystem": "Web",
  "description": "Reporting software for small ecommerce teams.",
  "publisher": {
    "@type": "Organization",
    "name": "Northwind Labs",
    "url": "https://www.example.com"
  },
  "offers": {
    "@type": "Offer",
    "name": "Starter plan",
    "price": "29.00",
    "priceCurrency": "USD",
    "url": "https://www.example.com/pricing"
  }
}
</script>

Before you add it everywhere, keep your expectations realistic. Google says structured data isn't required for its AI features and that there is no special markup for them. Ahrefs tracked 1,885 pages that added JSON-LD against roughly 4,000 matched control pages, and found no meaningful lift in citations on ChatGPT, AI Mode or AI Overviews. Pages that AI cites are more likely to carry JSON-LD, but that is a correlation, and sites that bother with markup tend to be well maintained in other ways too.

So treat JSON-LD as good housekeeping, not a ranking trick. It still helps with rich results in traditional search and with identifying your brand as a distinct entity. Three habits make it worth the effort:

The values must match what visitors see on the page. If your markup says $29 and your page says $39, you have just created another conflicting source.

The markup should be in the HTML your server sends, not added later by JavaScript or a tag manager. Crawlers that don't run scripts will never see it otherwise.

Keep it current. Update it whenever prices or plans change, or it becomes one more outdated source.

Don't confuse visibility with revenue

Mentions and citations show exposure. They don't show sales.

Track visibility next to conversions, sales conversations and customer feedback. Customer surveys that ask how people found you are imperfect but useful. Treat a citation as an early signal, nothing more.

Where to start

Begin with access: review your robots.txt, CDN and firewall, and confirm your key content loads without JavaScript. Then correct the facts you control, and add accurate JSON-LD to the pages where those facts live. After that, build a balanced set of questions from real search queries and customer conversations, and trace any repeated claims back to their sources. Use outreach where you can reach the author and new content where there's a real gap. Finally, report visibility alongside business results, and say plainly what the data can't tell you.