Teardown: Why a Rival's Programmatic Glossary Beat The Trade Desk's Into the AI Overview
Objective: Given two competing 'What Is Programmatic Advertising?' page specimens, only one of which is cited in Google's AI Overview, identify every real GEO defect in the uncited page without flagging cosmetic non-issues.
You're on The Trade Desk's content team. Google's AI Overview for 'what is programmatic advertising' cites a smaller ad-tech blog's glossary page instead of The Trade Desk's own page, even though The Trade Desk's page ranks higher organically and has more backlinks.
Compare the two page specimens and the brand-mention snapshot, flag every real GEO defect, and map each one to the lesson concept it violates.
Which of the losing page's real defects, across structure, credibility signals, and off-page brand mentions, actually explain why a smaller competitor won the AI Overview citation?
Before you start
What you'll need
- —Familiarity with the GEO concepts: answer-first structure, structured formatting, and credibility signals
- —Basic understanding of what schema markup and E-E-A-T signals are
- E-E-A-T
- Experience, Expertise, Authoritativeness, Trustworthiness, Google's framework for judging content credibility, reflected here in named authorship and cited dates.
- Brand mentions
- unlinked references to a company's name across the web (forums, reviews, social), distinct from backlinks and shown by recent data to correlate more strongly with AI citation.
Free path (everything below is enough to finish)
Free and sufficient for a single-page comparison and defect log
Free, first-party data on whether the page is even eligible to be crawled and re-evaluated
Paid upgrades (optional, faster/deeper)
Faster than manually compiling backlink counts, but not required to complete the teardown
The process
Specimens to review
Below are the opening sections of two 'What Is Programmatic Advertising?' pages. Only Page B is cited in Google's AI Overview for the query. Identify every GEO defect in Page A (The Trade Desk's version).
=== PAGE A: The Trade Desk (NOT cited) ===
H1: Understanding the World of Programmatic
By: Marketing Team
Advertising has changed enormously over the past two decades. What began as a handshake business between media buyers and publishers has evolved into a complex, automated ecosystem that operates in fractions of a second. Programmatic is at the center of that shift, and understanding it requires looking at where the industry has been.
## Why It Matters
Marketers today can't afford to ignore automation, and the numbers back that up...
[No structured data present in page source. No dateModified visible.]
=== PAGE B: Competitor (cited in AI Overview) ===
H1: What Is Programmatic Advertising? (Definition + How It Works)
By: Jane Torres, Head of AdTech Research — Updated March 2026
Programmatic advertising is the automated buying and selling of digital ad space using software and real-time bidding, instead of manual insertion orders.
## How It Works (4 Steps)
1. An ad exchange auctions available impressions in real time.
2. A demand-side platform bids on behalf of the advertiser.
3. The winning bid's creative is served instantly.
4. Performance data flows back for the next auction.
[Page source contains: <script type="application/ld+json">{"@type":"Article","author":"Jane Torres","datePublished":"2026-03-02"}</script>]Specimen: synthetic, realistic
The Trade Desk's SEO team also pulled a 90-day brand-mention snapshot for both companies. Identify the GEO defect this snapshot reveals.
THE TRADE DESK, 90-DAY SNAPSHOT New backlinks: 340 Reddit/forum mentions of "The Trade Desk": 4 Review-site mentions: 2 COMPETITOR, 90-DAY SNAPSHOT New backlinks: 85 Reddit/forum mentions: 61 Review-site mentions: 28
Specimen: synthetic, realistic
Analyze your findings
What to look for
- Opening definition
- Does the H1 and opening paragraph state a direct definition, or build up to one through history/context?
- Authorship signal
- Is there a named author with credentials, or a generic team byline?
- Structured data
- Is there JSON-LD schema and a visible publish/update date?
- Formatting
- Is a process explained as a numbered list, or as flowing prose?
- Brand mentions vs. backlinks
- Which signal is actually larger, and which one does the data show correlates more strongly with citation?
Make the call
The Trade Desk has 4x more backlinks than the competitor but 15x fewer Reddit/forum mentions. Both companies rank similarly well organically. What should the team prioritize?
Recommendation · Priority: High
“Fix Page A's four content-level defects first (add a direct opening definition, a named author byline, Article schema with a visible date, and convert the process explanation to a numbered list), all achievable within a sprint. In parallel, start a brand-mention campaign (contributing to relevant forum threads, seeking review-site presence) since it's the single highest-leverage, most under-invested signal the data reveals.”
Common mistakes
What trips people up
Flagging shorter H1s, URL slugs, or outbound-link counts as defects — these are the specimen's built-in distractors; none of them affect AI extraction or citation.
Assuming more backlinks should mean better AI visibility — the brand-mention snapshot shows the opposite pattern here, more backlinks did not prevent the citation loss.
Missing the schema/structured-data defect because the page 'looks fine' visually — JSON-LD and dateModified are invisible in the rendered page but are exactly what AI systems check for machine-readable freshness and authorship.
Treating the brand-mentions defect as separate from the content-structure defects instead of prioritizing it — it's the least visible defect in the specimens but the one the data shows matters most, deprioritizing it under-serves the actual fix.
Final deliverable
A defect log flagging every real GEO issue in Page A, each mapped to a lesson concept, plus a prioritized fix list.
See a reference example
Applying the same teardown to Ola Electric's 'What Is Regenerative Braking?' page: 3 critical defects found (buried definition, no schema, prose-only steps) and 1 moderate defect (no named author). Fix priority: rewrite opener (30 min) > add Article schema (dev ticket) > convert steps to a numbered list (30 min).
Success criteria
You're done when you can:
- Flags all 5 real defects across both specimens
- Does not flag any of the 5 distractors as defects
- Correctly identifies the brand-mentions defect as the least obvious but highest-leverage fix
Key takeaway
A losing page can have real defects at multiple levels: buried definitions, missing schema, prose instead of structured lists, and a weaker off-page signal profile. Identifying every real defect while correctly rejecting cosmetic distractors, then weighting the fix list by which signal the data shows matters most, is what separates a useful teardown from a surface-level content critique.