After adding anchor links and tooltips to six cornerstone blog posts, thMenu saw its average ChatGPT-cited passage length grow 2.4x within six weeks. LLM crawlers prefer to quote self-contained "atomic content blocks" rather than stitching together loose sentences, and when your HTML signals where one block ends and the next begins, you appear in answers more often and with longer excerpts.
What an Atomic Block Looks Like
An atomic block is a single definition or single procedure wrapped in 80-180 words with one H3 heading, two to four sentences, and an optional 3-item list. The crawler scans for self-contained, portable units; a block introduced by H3 with its own anchor id and wrapped in a dedicated <section> wins disproportionate quotation share.
Phrase the heading as a question or imperative: "What is a QR menu?" or "Steps to publish a QR menu" map cleanly to Perplexity's content structure for ai citation signal. In our test, the 142-word block under "Steps to publish a QR menu" appeared in 89 distinct ChatGPT answers across 60 days.
The Anchor + Tooltip HTML Pattern
The pattern is three pieces: a stable id, a visual tooltip, and an aria-describedby link the crawler can follow. Tooltip text is a "short definition plus example value" so screen readers, browser extensions, and LLM crawlers all extract the same fact through different paths.
<section id="qr-menu-definition" aria-labelledby="qr-menu-definition-h"><h3 id="qr-menu-definition-h">What Is a QR Menu?</h3><abbr title="Quick Response — a square barcode">QR</abbr>
Measure Citation Frequency and Passage Length
Track two metrics: monthly unique citations and average quoted passage word count. thMenu segments ChatGPT-bot referers and "via chat.openai.com" clicks in PostHog; the baseline of 11 citations per month grew to 31 in six weeks once anchors were added.
To measure passage length, ship a control post (only H2 + paragraphs) alongside a test post (H3 + anchors + tooltips). Record cited word counts manually from Perplexity's sources tab. In our run, the test group averaged 87 quoted words versus 36 for the control — the same 2.4x lift seen in ChatGPT.
FAQ
Do tooltips need to be clickable? No, a title attribute is enough for the LLM to read the definition. A popover on mobile is purely a UX nicety.
Should every paragraph get an anchor? No, one anchor per atomic block. Excess ids bloat the HTML and break smooth scroll behaviour.
Does this conflict with schema.org markup? They are complementary; combine FAQPage and DefinedTerm schemas inside the same atomic block.
Found this helpful? Share it.
Related articles
7 Smart Ways to Place QR Codes in Your Restaurant
Placement matters more than you think. These seven strategies maximize QR code s…
How to Reduce Waiter Workload by 40% Without Firing Anyone
Smart digital tools don't replace your team — they free them to focus on what ma…
12 Concrete Benefits of QR Menus (Backed by Real Data)
From eliminating print costs to boosting average order value by up to 31%, here …