Case study: Gumlet turned ChatGPT mentions into 20% of inbound revenue. Read it →
The Content Stack Test: every B2B SaaS piece needs to pass 4 retrieval tests before publishing.
Across the DerivateX 100-Piece Content Stack Audit, B2B SaaS content scoring 80 or higher on the unified Stack Score drove 4.7× more AI citations over 6 months than content scoring 60 to 80. That is not a content quality gap. It is a quality assurance gap.
DerivateX 100-Piece Content Stack Audit · 6-Month Citation Outcomes
80+ Stack Score
Top-tier content. Compounds across surfaces. Drives demos.
4.7×
citation lift
60–80 Stack Score
Publishable but plateaued. Occasional citation, no compounding.
TL;DR
The framework in six lines.
Every B2B SaaS marketing lead running content QA in 2026 should be able to repeat these six points. One platform isn't a strategy.
01
Modern B2B SaaS content has to pass four separate retrieval tests: Google rank-ability, ChatGPT extract-ability, Perplexity source-ability, Reddit relate-ability.
02
The Stack Score is 0 to 100, with 25 points per test and 16 binary checks total. Below 60: revise. 60 to 80: publish with weaknesses. 80+: top-tier.
03
Across the audit, 80+ pieces drove 4.7× more citations over 6 months than 60 to 80 pieces. The gap is not in length, links, or polish. It is in passing all four tests.
04
The most under-reported finding: high-Google low-ChatGPT pieces produce the worst 6-month performance. Buyers click, then bounce after cross-checking in AI search.
05
The full Content Stack Test runs in 30 minutes per piece once the workflow is set up. 6 to 8 minutes per test, plus 2 minutes for scoring.
06
If your content QA today is "Surfer score above 80, ship it," you are running a single-surface check against a four-surface buyer journey. The cost shows up in 6 months.
The Problem
One test isn't enough. Buyers research across four surfaces.
The standard B2B SaaS content QA workflow in 2026 looks like this. Writer drafts the piece. Editor checks for grammar and structure. Someone runs it through Surfer or Clearscope. The score comes back at 87. Ship.
That workflow was correct in 2020. It is wrong in 2026. The reason it is wrong is that the buyer journey now spans four retrieval surfaces, and a tool that scores one of them tells you nothing about the other three.
A typical $5M+ ARR B2B SaaS buyer in 2026 starts research on Google, sees the SERP and the AI Overview, asks ChatGPT for a shortlist, asks Perplexity to compare the top three with sources, and goes to Reddit for unfiltered opinions before booking a demo. By the time they enter your funnel, they have crossed all four surfaces.
A piece that performs on one surface and fails on three has missed three of four buyer touchpoints.
The Content Stack Test is the pre-publishing QA framework that catches this before the piece goes live. Four separate tests, scored independently, summed to a single 100-point Stack Score with named publish thresholds. Run it on every piece. Ship the ones that pass. Revise the ones that don't.
The Counterfactual
The bar that surprises everyone is at the top.
In our analysis, the worst 6-month performance came not from pieces that failed every test, but from pieces with high Google scores and low ChatGPT scores. Buyers clicked, then bounced after cross-checking in AI search. The Google rank became a liability, not an asset. Performance ranked from worst to best:
01
High Google, low ChatGPT
Surfer-optimized pieces that rank, then lose the buyer at the AI cross-check. Worst outcome.
~0.2×
02
Failed both tests
Pieces that bomb both Google and ChatGPT. Get neither the click nor the bounce.
~0.4×
03
Balanced 60 to 80 Stack Score
Adequate on most surfaces, weak on one or two. Occasional citation, no compounding.
1.0×
04
80+ Stack Score
Top-tier across all four surfaces. Compounds. Drives demos.
4.7×
Single-platform optimization is not a partial win. It is structurally worse than no optimization at all when the partial win is on Google and the gap is on ChatGPT.
The Architecture
Four tests. Sixteen binary checks.
Each test is worth 25 points, with 6.25 points per binary check. A piece passing all four checks in a test scores 25. Pass three, score 18.75. Pass two, 12.5. The full Stack Score is the sum across all four tests.
01
The Google Test
Rank-ability · Foundation, not finish line
25 pts
Word count matches top-3 SERP average plus 20%
Answers every People Also Ask question for the primary query
Has correct schema markup for the asset type
Has a link-pitchable hook: defensible angle, original data, or quote-worthy claim
02
The ChatGPT Test
Extract-ability · Most B2B SaaS content fails here
25 pts
Contains 3 to 5 self-contained 200-300 word chunks
Each chunk has at least one named entity as attribution anchor
H3s shaped as questions an ICP buyer would type, not keyword variations
FAQ has 4+ questions answered in 60 to 120 word self-contained form
03
The Perplexity Test
Source-ability · Different signals than ChatGPT
25 pts
Cites at least one 2025+ source. Recency weights heavily in live retrieval.
Claims attributed to specific named sources, not "experts say"
Contains at least one piece of original research, data, or proprietary observation
Page loads in under 2.5 seconds (Largest Contentful Paint)
04
The Reddit Test
Relate-ability · Hardest to fake, easiest to fix
25 pts
Uses first-person ICP language ("we tried this", "our team found")
Acknowledges the most common ICP objection on the topic
Contains at least one "I tried this and it didn't work" honest moment
Condensed to 200 words, would pass as a top-voted Reddit comment
In Action · Live Stack Scorer
Watch the Stack Score assemble in real time.
Three real B2B SaaS content scenarios from the audit. Click a scenario and watch each test run, each binary check resolve, and the unified Stack Score lock in. Notice that the Surfer-optimized blog scores worse than the Practitioner Post, despite winning Google.
0/100
Stack Score
Click a scenario to run
Google Test
0/25
Word count
PAA covered
Schema
Link-pitchable
ChatGPT Test
0/25
Chunked structure
Named entities
Question H3s
FAQ extractable
Perplexity Test
0/25
2025+ sources
Named attribution
Original data
Page speed
Reddit Test
0/25
First-person voice
Acknowledges objection
"Didn't work" moment
Reddit-grade authentic
The Thresholds
Three publish bands. Derived from the audit.
The publish thresholds are not soft guidance. They are derived from the citation outcome distribution in the 100-Piece Audit. The single biggest decision the Stack Score forces is the one most teams currently avoid: pieces below 60 should not be published.
< 60
Don't Publish
Revise before shipping
The piece will compound negative signals over its 6-month performance window. In the audit, pieces below 60 produced negative ROI relative to the time invested writing them.
Revising costs less than publishing, watching the piece underperform, and trying to retrofit it later.
60–80
Publish, with caveats
Document the gap, plan the fix
Most B2B SaaS content lives here. The value isn't refusing to publish. The value is knowing what the piece can't do, planning the fix, and not being surprised when it underperforms top-tier pieces.
Have a 30-day plan to address the weakness post-publication. Link-build, FAQ-rewrite, original-data appendix, or first-person rewrite.
80+
Top-Tier Content
Ship and let it compound
This is the band that produced the 4.7× citation lift in the audit. Pieces here drive demo bookings, get cited across all four LLMs, and compound across surfaces over their 6-month window.
Ship it. Track citation outcomes biweekly. Build more like this.
The 30-minute workflow
Once calibrated, the test runs in 30 minutes flat per piece
30m
7 min
Google Test: SERP word count, PAA, schema, link-pitchable angle
8 min
ChatGPT Test: chunks, named entities, question H3s, FAQ extractability
8 min
Perplexity Test: 2025+ sources, named attribution, original data, page speed
7 min
Reddit Test: first-person, objection, "didn't work" moment, gut-check
The System
The Stack Test sits inside a five-framework system.
The Stack Test does not exist in isolation. It is the QA layer that confirms the work upstream actually shipped. Click any framework below to see how the system fits together.
Step 01
Visibility Vacuum Theory
When to invest, by category stage.
Step 02
Search Budget
Where the spend goes, by retrieval surface.
Step 03
Citation Surface Map
What shape the surface needs, layer by layer.
Step 04
Citation Engineering
How each piece gets engineered.
Step 05
Content Stack Test
Whether the engineering actually worked, before publishing.
Visibility Vacuum tells you when. Search Budget tells you where. Citation Surface Map tells you what shape. Citation Engineering tells you how. The Content Stack Test tells you whether you actually got it right before publishing.
The Diagnostic
Where you are right now.
After reading this, you are in one of three positions on your existing content workflow. The audit data is clear: teams that adopt the Stack Test as a formal pre-publish gate produce higher-citation content within 6 months. Teams that don't keep shipping pieces that compound the wrong signals.
Position 01
No pre-publish QA beyond editorial
Content scores high on grammar and structure and ships. You have no idea which surfaces it performs on. Most teams in this position discover they have been shipping 50 to 65 Stack Score content for months.
The Move
Run the Stack Test on the next 5 pieces in your queue. Just to see the score distribution.
Position 02
Surfer or Clearscope, ship
Single-surface QA scores Google rank-ability. The other three surfaces are ungoverned. You are the team most exposed to the high-Google low-ChatGPT failure mode the audit identified.
The Move
Layer ChatGPT, Perplexity, and Reddit Tests onto your existing workflow. 30 minutes per piece.
Position 03
Informal multi-surface QA
You think you're already doing this. Most teams in this position are running an inconsistent version of one or two of the four tests. Different reviewers score differently, different pieces ship under different rubrics.
The Move
Formalize the Stack Score. Same scoring, same thresholds, same revise/ship rubric every piece.
FAQ
Common questions from operators.
The Content Stack Test is a pre-publishing QA framework for B2B SaaS content that scores every piece against four separate retrieval tests before it goes live: the Google Test (rank-ability), the ChatGPT Test (extract-ability), the Perplexity Test (source-ability), and the Reddit Test (relate-ability). Each test has 4 binary checks and is worth 25 points, summing to a unified 0-100 Stack Score. The framework replaces single-platform content QA workflows like Surfer or Clearscope with multi-surface QA that scores how a piece will perform across all four surfaces a B2B SaaS buyer crosses before purchase.
The Stack Score is a 0-100 unified pre-publishing metric that sums the four Content Stack Test scores. Each test contributes 25 points, with 6.25 points per binary check across 16 total checks. To calculate it, run all four tests on the piece, score each binary check as pass (6.25 points) or fail (0 points), and sum the totals. A score below 60 means don't publish, revise. A score of 60 to 80 means publish with documented weaknesses and a 30-day fix plan. A score of 80 or higher means top-tier content, the band that produced 4.7x citation lift in the DerivateX 100-Piece Content Stack Audit.
Surfer and Clearscope score one platform: Google. They are excellent at predicting Google rank based on SERP correlation. They score zero of the other three retrieval surfaces a B2B SaaS buyer crosses. In our analysis, content with high Google Test scores but low ChatGPT Test scores consistently produces the worst 6-month performance. Buyers find these pieces, click through, then bounce after cross-checking in ChatGPT and seeing better-cited competitors. Single-platform optimization is not a partial win. It is structurally worse than no optimization when the partial win is on Google and the gap is on ChatGPT.
Between Q4 2024 and Q4 2025, the DerivateX content team published 100 pieces of original B2B SaaS content across 8 client engagements. Every piece was scored on the four Content Stack Tests at publication. Citation outcomes were tracked biweekly across ChatGPT, Perplexity, Claude, and Gemini for 6 months following each publication. The headline finding: pieces scoring 80 or higher on the Stack Score drove 4.7x more cumulative citations than pieces scoring 60 to 80. Top-performing pieces were not the longest or the most heavily backlinked; they were the pieces that passed all four tests.
Open the draft in one tab. Open ChatGPT, Perplexity, and a Reddit-aware authenticity check tool in others. Run the Google Test in 7 minutes (top-3 SERP word count, PAA coverage, schema, link-pitchability of the angle). Run the ChatGPT Test in 8 minutes (3 to 5 self-contained chunks, named entities per chunk, question-shaped H3s, FAQ extractability). Run the Perplexity Test in 8 minutes (2025+ sources, named attribution, original data, page speed). Run the Reddit Test in 7 minutes (first-person language, objection acknowledgment, honest moments, 200-word Reddit comment gut check). Sum the four scores and apply the publish threshold. Our free AEO Content Evaluator automates much of the ChatGPT and Perplexity checks.
The Action
One platform isn't a strategy. Stop publishing single-surface content.
Run the Content Stack Test on the next piece in your publication queue, before it goes live. The 30-minute investment is the difference between content that compounds and content that doesn't.
01
Pick the next piece in your queue
The one your team was about to ship after the editorial review. Don't pick a published piece. Pick the next unshipped one.
02
Run all four tests in 30 minutes
16 binary checks. Pass or fail. No partial credit. Sum the scores. Get a Stack Score. Apply the threshold.
03
Hold the line on the threshold
If the piece scores below 60, delay publication and revise. The right call is uncomfortable in the moment and right in retrospect every single time.
Related Reading
Connected frameworks and data.
Four platforms. Four tests. One Stack Score. Stop publishing content that's only ready for Google.