Unclear primary call-to-action

Agents can't tell what action your page wants users to take, so they don't recommend it for buying or signup queries.

The primary call-to-action (CTA) is the single highest-intent thing a visitor can do on a page: buy this product, start a free trial, book a demo, sign up, download the app. "Unclear primary CTA" means the page either has no obvious CTA, has many competing CTAs of equal weight, or buries the CTA inside ambiguous copy. From an agent's perspective, this looks like a page with no purpose, and agents prefer pages with crisp purposes when answering action-oriented queries.

AI shopping agents and task agents are increasingly the funnel: a user asks Claude or ChatGPT 'help me find a CRM and start a trial' and the agent picks one. To pick yours, the agent needs to extract the next action ("start a free trial," "book a demo," "buy now") from your page in plain text. Pages with one prominent CTA in semantic HTML (a real <a> or <button> with descriptive text) win. Pages where the CTA is a styled <div> with onClick handlers, or where every section has its own competing button, lose.

For commerce: agents that can identify your "Add to cart" or "Buy now" button in the rendered HTML are eligible for Agentic Commerce flows (Stripe Agent Pay, OpenAI shopping). Pages that hide the action behind a JavaScript-rendered modal are not. For SaaS: agents recommending tools embed the CTA URL directly in the answer; if the agent cannot identify the trial-start URL, your competitor's answer wins. The fix is mostly a copy-and-markup audit, not infrastructure.

Open your most-trafficked landing page. Read the first viewport (no scrolling) and ask: what is the one thing this page wants me to do? If you can't answer in five words, the CTA is unclear. Then view-source and confirm the CTA is a real <a href="…"> or <button> with descriptive text ("Start free trial," "Add to cart," "Book a demo"). If the action label is generic ("Click here," "Learn more"), agents can't map it to a user intent.

Pick one primary CTA per page. Make it a real anchor or button with text that names the action explicitly. Place it above the fold, repeat it after the value proposition, and avoid putting four equally-styled secondary CTAs (Twitter, GitHub, Docs, Pricing) at the same visual weight. If the page has multiple legitimate paths (sign up vs. learn more), use visual hierarchy: one dominant button, one ghost-style secondary.

<header>
  <h1>Stop manually copying expense receipts.</h1>
  <p class="lede">Acme reads your inbox, categorises expenses, and hands them
     to your accountant. Free for the first 30 days.</p>

  <!-- Primary CTA: descriptive label, real <a> with destination URL,
       visually dominant. -->
  <a href="/signup" class="btn btn-primary">Start your free trial</a>

  <!-- Secondary CTA: lower visual weight, no competition with primary. -->
  <a href="/demo" class="btn btn-ghost">Book a 15-minute demo</a>
</header>

The reference snippet above is platform-agnostic. Below are the concrete steps for the four most common stacks plus a generic fallback for any other host.

import Link from 'next/link';

export default function Home() {
  return (
    <header>
      <h1>Stop manually copying expense receipts.</h1>
      <p>Acme reads your inbox and hands expenses to your accountant.</p>

      {/* Server-rendered, in initial HTML, agent-extractable. */}
      <Link href="/signup" className="btn btn-primary">
        Start your free trial
      </Link>
      <Link href="/demo" className="btn btn-ghost">
        Book a demo
      </Link>
    </header>
  );
}

<!-- Wrong: opens via JS, agents can't extract destination. -->
<div class="btn-primary" onclick="window.location='/signup'">Start trial</div>

<!-- Right: real anchor, agent-readable, works without JS. -->
<a class="btn-primary" href="/signup">Start your free trial</a>

<form method="post" action="/cart/add">
  <input type="hidden" name="id" value="{{ product.selected_or_first_available_variant.id }}">

  <!-- Real <button type="submit"> — agents can identify it as the
       primary action and AI shopping agents can fill it. -->
  <button type="submit" class="btn-primary">
    Add to cart — {{ product.price | money }}
  </button>
</form>

Hero CTA → Link Block element with "btn-primary" class
Secondary CTA → Link Block, "btn-ghost" class, lower visual weight
Footer CTAs → Link Block, smallest visual weight

Avoid: Div Block with click interaction. The interaction works in
the browser but gives agents no destination URL.

<section class="hero">
  <h1>Stop copying receipts.</h1>
  <p>Acme reads your inbox and categorises expenses automatically.</p>

  <a href="/signup" class="btn-primary">Start your free trial</a>
  <a href="/demo" class="btn-ghost">Book a 15-minute demo</a>
</section>

• Five primary buttons of equal weight in the hero ("Sign up," "Watch demo," "Read docs," "See pricing," "Read blog"). Pick one. Demote the rest visually or move them to the nav.
• Generic CTA labels ("Click here," "Learn more," "Submit"). Agents map labels to intents. "Start your free trial" is an intent; "Submit" is not.
• Styling a <div> as a button. The page looks fine to humans; agents see a div with no destination. Use <a> for navigation, <button> for actions.
• JS-only CTAs that don't have an href. Even if the click handler does the right thing, agents that don't run JS see a dead element. Provide the href as a fallback.
• Hiding the CTA behind a modal that opens on a tiny "Get started" link. Agents extract the visible primary action; modals are invisible to most crawlers.

View source on your top three pages and confirm exactly one prominent <a> or <button> with descriptive label appears in the first viewport. Run a fresh AgentSpeed scan, and the actionability category score will tick up. Test the page through an AI shopping agent or by asking ChatGPT/Claude "what action should I take on this page?"; the answer should match your intended primary CTA.