The Answer Unit Teardown: Grading 5 Real AI-Cited Passages
Objective: Given 5 specimen passages that won or lost an AI citation, correctly diagnose which structural defect caused each outcome using the Answer Unit framework and the freshness and schema signals from the lesson.
You're auditing Peloton's blog and support content ahead of a content refresh sprint. Your manager wants a defect list, not just a rewrite, so the writing team understands exactly what wins and loses AI citations.
Score 5 specimen passages against the Answer Unit framework, flag the specific defect in each losing passage, and don't get fooled by the 2 distractor prompts per item that look like defects but aren't.
For each losing passage, which specific defect (buried answer, stale data, missing schema, or an authority gap that no rewrite fixes) caused the citation loss?
Before you start
What you'll need
- —Completion of, or familiarity with, the Answer Unit framework: a question heading, a 40-60 word direct answer, then supporting detail
- —Understanding that freshness and schema are separate signals from content structure
- Answer Unit
- the framework of a question-phrased heading immediately followed by a 40-60 word direct answer, with supporting detail placed after it, not before.
- Schema markup
- structured, machine-readable code (like FAQPage schema) added to a page that explicitly labels content, such as a Q&A pair, for crawlers.
Free path (everything below is enough to finish)
Free, no account friction
Free tier confirms the diagnosis against live behavior
The process
Specimens to review
This passage ranks #4 on Google for 'peloton bike plus rotating screen' but is never cited in AI Overviews or ChatGPT answers for that query. Identify the structural defect.
H2: 'Everything to Know About the Peloton Bike Plus Rotating Screen' Peloton has always been at the forefront of connected fitness innovation, bringing together world-class instructors and cutting-edge hardware. Since the original Bike launched in 2014, the company has continued to refine the at-home cycling experience for millions of members worldwide. The Bike Plus builds on this legacy with several thoughtful upgrades designed around member feedback. One of the most requested upgrades was screen rotation, which lets members pivot the 23.8-inch HD touchscreen 180 degrees to follow along with floor-based strength and stretching classes without leaving the bike.
Specimen: synthetic, realistic
This page used to win the featured snippet for 'peloton membership cost' but stopped appearing in AI Overviews eighteen months ago. What's the defect?
H2: 'How Much Does a Peloton Membership Cost?' A Peloton All-Access Membership costs $44 per month (data from our 2021 pricing page) and includes unlimited classes across Bike, Tread, and Row equipment for the whole household.
Specimen: synthetic, realistic
This passage is structurally a clean 40-60 word answer unit under a matching question heading, no fluff, no staleness. Yet a smaller fitness blog with the identical structure wins the citation instead. Is this a structural defect?
H2: 'What Is the Peloton Output Bike Metric?' Output is Peloton's blended score for how hard you worked in a class, combining your cadence and resistance to estimate power in watts and total calories burned. It updates in real time on your Bike or Bike Plus screen so you can pace effort against your own past rides.
Specimen: synthetic, realistic
This is a clean answer unit on an FAQ page with 12 other Q&A pairs formatted identically, but none of them carry FAQPage schema. Google's featured snippet shows the passage; ChatGPT never cites it. What's the fix?
H2: 'What Cadence Should I Ride at for a Peloton Beginner Class?' Beginners should aim for a cadence of 60 to 80 RPM at low-to-moderate resistance, which builds cardio endurance without overloading the knees. This range is called out directly by instructors in every Peloton beginner-labeled class.
Specimen: synthetic, realistic
The actual yes/no answer to 'is it worth it for small apartments' is buried after 5 unrelated spec bullets. What's the fix, beyond just deleting the bullets?
H2: 'Is the Peloton Row Worth It for Small Apartments?' - Weighs 145 lbs, foldable footprint - Costs $3,195 plus membership - Includes a 23.8-inch swivel screen - Offers over 20 workout types - Ships fully assembled in most metros For apartment dwellers with limited space, the Peloton Row folds vertically to roughly 20 inches wide when stored, making it one of the more compact rowers in its price class.
Specimen: synthetic, realistic
Analyze your findings
What to look for
- Answer position
- Does the direct answer appear in the first sentence under the heading, or after unrelated context?
- Freshness signals
- Does the passage cite a date, price, or figure that looks outdated with no visible update?
- Schema presence
- Is the Q&A wrapped in machine-readable schema (e.g. FAQPage), or is it plain text only?
- Structural vs. authority cause
- Is the passage already a clean Answer Unit, meaning the loss might be a domain-authority gap instead?
- Distractor plausibility
- Does a suggested defect actually break extraction, or is it a cosmetic detail (bolding, list style) that doesn't matter?
Make the call
Item 3 (the Output metric passage) is a clean 40-60 word Answer Unit under a matching question heading, with no staleness and no missing schema, yet a smaller blog with an identically structured passage wins the citation. What should the writing team do?
Recommendation · Priority: Medium
“Fix items 1, 2, 4, and 5 in the next content sprint: move buried answers to the first sentence, refresh the stale price with a visible update date, and add FAQPage schema to the cadence FAQ page. Do not assign a rewrite to item 3, flag it for the authority/PR team instead, since the Answer Unit framework has nothing left to fix there.”
Common mistakes
What trips people up
Assuming every citation loss has a content-structure cause — item 3 proves a passage can be structurally perfect and still lose on authority; misdiagnosing it wastes a rewrite that won't change the outcome.
Flagging cosmetic details (bolding, list format, hyperlinks) as the defect — these don't affect what an answer engine extracts; the distractors in each item exist specifically to test this.
Missing a stale-data signal because the content otherwise reads as high-quality — a visibly outdated price or figure is deprioritized even when the surrounding writing and structure are strong.
Treating a Google featured snippet win as proof AI engines will cite the same passage — item 4 shows a passage can win a Google snippet on structure alone while still losing an AI citation to a missing schema signal.
Final deliverable
A defect log for all 5 specimens: defect, severity, and a one-line fix recommendation, with item 3 correctly flagged as a non-structural authority gap.
See a reference example
Yelp answer-unit defect log (excerpt) Item | Verdict | Defect | Fix 1 | Buried answer | Direct answer arrives after 80 words of brand history | Move the 1-sentence definition to the first sentence 2 | Stale content | 2021 price with no update | Refresh price + add visible last-updated date 3 | No defect (authority gap) | N/A | Build backlinks/mentions, not a rewrite 4 | Missing schema | No FAQPage schema | Add FAQPage schema wrapping the Q&A 5 | Buried answer | Specs list precedes the yes/no verdict | Lead with verdict, move specs below
Success criteria
You're done when you can:
- Correctly diagnoses at least 4 of 5 specimen defects
- Correctly identifies item 3 as a non-structural (authority) issue rather than misdiagnosing it as a content defect
- Correctly rejects at least 8 of the 10 distractor prompts across all 5 items
Key takeaway
Not every citation loss has the same cause: a buried answer, stale data, and missing schema are all fixable with a rewrite or a dev ticket, but a structurally clean passage losing to a smaller competitor is usually an authority gap that no amount of rewriting solves. Diagnosing which one you're looking at, before recommending a fix, is the actual skill.