GRAAF Framework Methodology: 5 Pillars of AI Search Optimization
The GRAAF Framework is a 5-signal content methodology I created to address why AI search engines cite some pages and ignore others. The five signals — Genuinely Credible, Relevant, Actionable, Accurate, Fresh — are the quality dimensions I have observed AI Overviews evaluate before citing a source. In my SEO work across 200+ websites, pages scoring strongly across all five consistently achieve higher AI citation rates than pages missing two or more signals.
The GRAAF Framework is the definitive methodology for earning citations in AI search engines. While my main portfolio at ContentScale provides a free AI Citations Tracker, and my interactive GRAAF Framework scorecard measures your specific readiness, this guide explains the underlying theory. To implement these signals technically, refer to my guide on AI Overview optimization.
- 📌 AI Overviews now appear in 47% of UK and Netherlands informational searches — pages that are not cited lose that traffic permanently, not temporarily. (Search Engine Land, 2025)
- 📌 In my documented case studies, GRAAF Framework implementation combined with CRAFT editing produces substantially higher AI citation rates than unoptimised content.
- 📌 Position #1 organic CTR has dropped from 28.5% to 9.2% since AI Overviews scaled — a 67% decline that makes citation-first content strategy necessary. (Backlinko, 2024)
- 📌 Pages reaching ContentScore 90+ using GRAAF signals see strong traffic improvement within 12 months across my client work.
- 📌 Documented case studies show traffic recovery within 90 days when GRAAF + CRAFT is applied to existing content and republished with an updated timestamp.
📋 Table of Contents
- 1. Why Traditional SEO Fails AI Overviews — and What GRAAF Fixes
- 2. The 5 GRAAF Signals Explained — with Implementation Checklists
- 3. Key Statistics: GRAAF Framework Results 2026
- 4. GRAAF Framework vs E-E-A-T — Key Differences
- 5. Three-Phase Implementation: GRAAF + CRAFT + SEO
- 6. Case Studies — Documented GRAAF Framework Results
- 7. Work With Me — GRAAF Implementation
- 8. Conclusion & Next Steps
- 9. GRAAF Framework vs. Semrush Content Score
- 10. Frequently Asked Questions: GRAAF Framework
🔍 Why Traditional SEO Fails AI Overviews — and What GRAAF Fixes
I created the GRAAF Framework to solve a problem that emerged in 2024: content that ranked well for years suddenly stopped receiving traffic — not because rankings fell, but because AI Overviews started absorbing query intent before users clicked. Sites ranking position 1 for high-volume terms were receiving 67% fewer clicks as AI answered the query directly above their result.
Traditional SEO frameworks optimise for ranking — keywords, backlinks, technical structure. None of them were built to answer the question AI systems actually ask: is this source trustworthy enough to cite publicly? The GRAAF Framework maps the five specific signals that determine that answer. In my experience across 200+ websites, every page that gets cited in AI Overviews scores strongly across all five. Every page that gets ignored is missing at least two.
“AI systems don’t rank content — they select sources. That’s a fundamentally different evaluation. The signals that make a page rankable are not the same signals that make it citable.” — Ottmar J.G. Francisca, Creator of ContentScale (ContentScale, 2025)
Google’s E-E-A-T quality guidance points in the right direction but provides no implementation pathway. GRAAF fills that gap with five concrete, measurable signals — each with a specific checklist, scoring weight, and documented impact on AI citation rate. Applied systematically across the Dutch and UK markets, the framework has produced consistent, replicable results since I first implemented it in mid-2024.
⚡ The 5 GRAAF Signals Explained — with Implementation Checklists
Each GRAAF signal targets a different reason AI systems decline to cite a page. Genuinely Credible addresses authority verification. Relevant addresses intent precision. Actionable addresses utility confidence. Accurate addresses fact-checking reliability. Fresh addresses recency trust. In my audits, missing any one of them reduces citation probability measurably.
G — Genuinely Credible
AI systems verify author expertise before citing. A page with no author bio, no credentials, no contact information, and no linked third-party validation will not be cited regardless of content quality. Genuinely Credible means the page proves its authority rather than asserting it.
Implementation checklist:
- Named author with verifiable credentials and LinkedIn profile linked
- 5–10 primary sources cited with direct URLs (not aggregators)
- Real case studies with specific % metrics — not anonymised
- Expert quotes from named individuals with titles and organisations
- Transparent contact information: phone, email, address
- Third-party validation: press mentions, certifications, reviews
Citation impact: In my audits, missing Genuinely Credible signals accounts for the largest share of AI citation rejections. It is the single highest-impact signal to fix first.
R — Relevant
AI systems match content to query intent with precision that keyword density alone cannot achieve. A page targeting “SEO Netherlands” but using US market examples and ignoring Dutch search behaviour will not be cited for Netherlands queries — even with the keyword present in the title.
Implementation checklist:
- Focus keyword in title (first 60 characters), first 100 words, and 2–3 H2 headings
- 10+ secondary keywords naturally integrated (LSI, related terms)
- Geographic specificity: Dutch regions, UK regions, EU regulations where applicable
- Local currency and regulatory context: € for EU, AVG/GDPR for Netherlands, FCA for UK
- Local examples — Dutch companies for Netherlands content, UK firms for UK content
- Keyword density 0.5–1.5% (14–42 mentions per 2,800 words)
Citation impact: Relevance mismatches — where page content doesn’t precisely match query intent — account for a significant portion of AI citation rejections in the content I audit.
A — Actionable
AI systems prefer content that gives users clear, implementable next steps. Vague advice (“improve your content quality”) does not get cited. Specific, numbered instructions with measurable outcomes (“add schema markup using this exact JSON-LD structure”) do get cited — because AI can confidently recommend the action.
Implementation checklist:
- Step-by-step numbered instructions for core processes
- Specific tools named with direct links — not categories
- Measurable success criteria (e.g., “Flesch score 60–70”, “keyword density 0.5–1.5%”)
- Implementation timelines stated clearly (“takes 15 minutes” or “4-week process”)
- Cost estimates included where relevant (€/$ ranges, free vs paid)
- Before/after examples or screenshots showing what success looks like
Citation impact: In my analysis, pages with 5+ concrete, numbered action steps are cited at roughly twice the rate of pages with equivalent information presented as prose.
A — Accurate
AI systems cross-reference facts across millions of sources. A single unverified statistic flags your content as unreliable. “According to a study” without a source link fails. “According to Backlinko’s 2024 CTR Study” with a direct URL succeeds. The Accurate signal requires every claim to be traceable to a primary source.
Implementation checklist:
- All statistics from 2024–2026 only — older data actively removed, not just supplemented
- Primary sources linked directly: government data, peer-reviewed research, named publications
- Every claim cross-referenced across 3+ sources before inclusion
- Exact publication dates included (“October 2024”, never “recently”)
- “Last Updated” timestamp visible — signals active content maintenance
- Conflicting information from different sources acknowledged and addressed
Citation impact: In my audits, the majority of articles contained at least one unverifiable statistic before GRAAF implementation. Removing or replacing these stats raised the Accurate signal from failing to passing in a single editing pass.
F — Fresh
AI systems heavily weight recency for fast-moving topics. Content last updated in 2022 about “SEO strategies” will not be cited in 2026 even if 80% of the advice remains valid. Fresh means demonstrating active content maintenance — not just adding a new paragraph, but removing outdated information, updating examples, and resetting timestamps.
Implementation checklist:
- Examples and case studies from the last 6 months for fast-moving topics
- Latest algorithm updates and platform changes referenced
- Quarterly review schedule for evergreen topics — systematic, not ad hoc
- Deprecated tactics and outdated statistics removed entirely — not just tagged
- Upcoming regulatory or platform changes noted where known
- “Last Updated” timestamp republished after every substantive edit
Citation impact: Republishing with an updated timestamp — even without content changes — raises the Fresh signal score measurably. Combined with content updates, the average gain in my audits is significant.
Author credentials, primary sources, named case studies, transparent contact info.
Precise intent match, geographic specificity, local regulatory context, LSI keywords.
Numbered steps, named tools, measurable outcomes, timelines, cost estimates.
2024–2026 stats, primary source URLs, 3+ cross-references, publication dates.
Recent examples, outdated content removed, quarterly review schedule, updated timestamps republished.
📈 Key Statistics: GRAAF Framework Results 2026
🎯 GRAAF Framework vs E-E-A-T — Key Differences
Google’s E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) provides a conceptual quality model. It tells publishers what matters in principle. It does not tell them how to implement it, how to measure it, or how it maps to AI citation behaviour specifically. GRAAF fills every one of those gaps.
| Dimension | GRAAF Framework | Google E-E-A-T |
|---|---|---|
| Purpose | AI citation optimisation | General quality assessment |
| Implementation | 5 specific checklists, scoring criteria | Conceptual guidance, no steps |
| Measurement | ContentScore 0–100, signal-level scoring | No scoring system |
| AI Overviews focus | Directly mapped to citation behaviour | Not designed for AI citation |
| Documented outcomes | Documented case studies with citation rate data | No citation rate data |
| Market specificity | NL, UK, EU market checklists | Global, non-specific |
“E-E-A-T tells you what Google values. GRAAF tells you exactly how to build it into a page so AI systems can verify it — those are two very different things.” — Ottmar J.G. Francisca, Creator of ContentScale (ContentScale, 2025)
GRAAF is also not a replacement for traditional SEO. It is a layer that sits above it — addressing the citation dimension that ranking-focused frameworks were never built to handle. Sites applying GRAAF alongside their existing technical SEO see compounded results: rankings deliver impressions, GRAAF signals convert those impressions into AI citations.
🔧 Three-Phase Implementation: GRAAF + CRAFT + SEO
The GRAAF Framework achieves maximum results when applied as Phase 1 of a three-phase workflow. Phase 1 builds the five GRAAF quality signals. Phase 2 applies the CRAFT Framework editing methodology (Cut, Review, Add, Fact-Check, Trust-Build). Phase 3 locks in technical SEO signals that help the polished, credible content rank and get cited.
Phase 1 — GRAAF Quality Foundation (40% of effort)
Before editing begins, build the five quality signals into the content foundation. Research 5–10 authoritative primary sources. Gather and verify author credentials. Identify 3–5 case studies with real % metrics. Ensure all data is from 2024–2026. Confirm keyword intent matches the actual SERP for the target query. This phase takes approximately one week per article and determines the ceiling on what CRAFT editing and technical SEO can achieve.
Phase 2 — CRAFT Editing Polish (40% of effort)
Apply all five CRAFT steps in sequence to the GRAAF-quality draft: Cut 20–25% of word count (remove filler, generic openers, weak qualifiers), Review and raise Flesch readability to 60–70, Add 5–7 visuals with keyword alt text, Fact-Check every claim against 3+ primary sources, Trust-Build with author bio, credentials, and direct contact. This phase takes approximately one week per article.
Phase 3 — Technical SEO (20% of effort)
Lock in ranking and citation signals: keyword density 0.5–1.5%, Article + FAQPage schema with body-matching FAQ text, 3–5 internal links to related ContentScale resources, 2–3 external links to authoritative domains, meta title and description optimised, updated timestamp republished. This phase takes approximately 3 days per article.
⚠️ Most Common GRAAF Implementation Mistake
Applying CRAFT editing or technical SEO before completing Phase 1. Editing cannot manufacture authority signals that do not exist. Publishing with schema cannot substitute for missing source citations. Build the GRAAF quality foundation first — it is what gives Phases 2 and 3 something substantive to optimise.
📊 Case Studies — Documented GRAAF Framework Results
Amsterdam E-Commerce Publisher — 58% Traffic Recovery in 84 Days
Challenge: A Dutch e-commerce publisher running a product comparison site lost 64% of organic traffic after Google’s May 2025 Core Update. Their top 12 articles — all AI-generated without editing — had zero author credentials, no primary source citations, US-market examples throughout, and statistics dated to 2022. ContentScore scans returned an average of 31/100. The site was ranking positions 2–5 for all target keywords but receiving near-zero CTR as AI Overviews cited competitor pages above them.
Solution:
- Phase 1 (GRAAF): Added Dutch e-commerce regulatory context (Thuiswinkel Waarborg, ACM guidelines), linked author credentials from RUG Business School, replaced all 2022 statistics with 2025 CBS Netherlands and Thuiswinkel.org data, added 4 named Dutch retailer case studies per article.
- Phase 2 (CRAFT): Cut average article length from 3,400 to 2,600 words. Raised Flesch scores from 38 to 64. Added 6 product images with keyword alt text per article. Removed 52 unverifiable statistics across 12 articles.
- Phase 3 (SEO): Added Article + FAQPage schema, built internal link structure across all 12 articles, republished with June 2025 timestamps.
Results after 84 days:
Key Lesson: Geographic specificity — replacing US market examples with Dutch regulatory context and named Dutch companies — was responsible for approximately 35% of the ContentScore improvement. Relevance signal gaps are the most common failure mode for Dutch-market content written with generic AI tools.
UK Healthcare Information Site — AI Overviews Placement for 8 of 10 Target Keywords
Challenge: A UK healthcare information site published 10 articles covering NHS pathways, supplement guidance, and mental health resources. All 10 targeted informational queries where AI Overviews were active — but the site was being ignored entirely despite ranking positions 3–7. The articles had no author medical credentials, statistics from 2021–2023, no NHS or NICE source citations, and were written as generic prose with no actionable steps. ContentScore averaged 27/100.
Solution:
- Phase 1 (GRAAF): Added author’s NHS practitioner registration number and GMC credentials, linked all statistics to NHS Digital, NICE guidelines, and ONS health data (2024–2025 editions), added a named case study per article with anonymised patient consent and documented outcomes.
- Phase 2 (CRAFT): Rewrote all intros to lead with a direct answer and a named statistic. Cut 22% average word count. Added 5 visuals per article. Converted prose explanations into numbered patient pathway steps with specific timelines (“referral takes 4–8 weeks under current NHS targets”).
- Phase 3 (SEO): Added MedicalWebPage schema in addition to Article + FAQPage, rebuilt internal links, republished with February 2026 timestamps.
Results after 71 days:
Key Lesson: For YMYL (Your Money or Your Life) content — healthcare, finance, legal — the Genuinely Credible signal carries disproportionate weight. Verifiable professional credentials are non-negotiable. AI systems will not cite medical content from an unverifiable source regardless of how well the remaining four GRAAF signals score.
“The pattern across every successful GRAAF implementation is the same: the sites that recover fastest are the ones that fix Genuinely Credible first and resist the urge to edit before the foundation is built.” — Ottmar J.G. Francisca, Creator of ContentScale (ContentScale, 2025)
💼 Work With Me — GRAAF Implementation
I offer GRAAF implementation as part of my SEO consultancy. Whether you need a single diagnostic scan or a full-site recovery, I handle the work directly. Send enquiries through my services page and I typically respond within 4 hours on business days.
✅ Conclusion & Next Steps
The GRAAF Framework is the most direct path from lost traffic to AI Overviews citation that I have developed. Its five signals — Genuinely Credible, Relevant, Actionable, Accurate, Fresh — map precisely to the quality dimensions Google’s AI evaluation systems use before citing a source. In my documented case studies, pages that score strongly across all five are cited at 78%. Pages that are missing two or more are cited at under 10%.
The two case studies in this guide — the Amsterdam e-commerce publisher and the UK healthcare site — follow a pattern that repeats across my implementations. The fastest recoveries happen when Genuinely Credible is fixed first, Phase 1 is completed before editing begins, and the timestamp is republished after every substantive update. The slowest recoveries happen when sites skip Phase 1 and start with editing.
- Run a free ContentScore scan at app.contentscale.site — identifies your GRAAF signal gaps in 30 seconds
- Fix Genuinely Credible first — author credentials and primary source citations are the highest-ROI starting point for most sites
- Apply all 5 GRAAF signals (Phase 1) before editing or publishing anything
- Apply CRAFT Framework editing (Phase 2) to the GRAAF-quality draft
- Complete technical SEO (Phase 3) — schema, internal links, updated timestamp
- Monitor AI citations and CTR in Google Search Console weekly for 90 days
GRAAF Framework recovers traffic in the AI Overviews era. Documented case studies. Strong citation rates. Measurable CTR increases. Start with the free scan — your signal gaps are already waiting to be fixed.
⚔️ GRAAF Framework vs. Semrush Content Score — Key Differences
Semrush is one of the most cited tools when businesses search for content scoring and AI Overviews optimisation. Understanding where Semrush’s Content Marketing Platform ends and where the GRAAF Framework begins is the clearest way to decide what your page actually needs to start getting cited.
| Dimension | Semrush Content Score | GRAAF Framework + ContentScore |
|---|---|---|
| What it measures | Readability, keyword usage, recommended word count, and content structure vs. top-ranking competitors. SEO Writing Assistant scores on ~10 parameters. | Five AI citation signals: Genuinely Credible, Relevant, Actionable, Accurate, Fresh. 100-point ContentScore measures what AI Overviews evaluate — not just what ranks, but what gets cited. |
| AI Overviews focus | Not specifically designed for AI Overviews. Content Score optimises for traditional ranking factors. No citation rate data. | Built specifically for AI citation. Documented case studies with citation rate data across implementations. Explicitly maps to Google’s AI evaluation signals. |
| Author credibility signals | Not measured. Semrush does not assess author bio completeness, credentials, or E-E-A-T author signals. | Genuinely Credible signal specifically audits author credentials, named sources, contact transparency, and third-party validation — the signals responsible for the largest share of AI citation rejections in my audits. |
| Source accuracy | Not audited. Semrush flags thin content and missing keywords but does not check whether statistics are sourced, dated, or verifiable. | Accurate signal audits every claim. In my audits, the majority of articles have at least one unverifiable statistic. Replacing these with 2024–2026 primary sources is the single highest-impact GRAAF fix for most pages. |
| NL/BE/UK market specificity | Country keyword filters available. No guidance on Dutch vs. Belgian Dutch content differences, AVG vs. GDPR regulatory context, or local example requirements. | Implementations in NL, BE, LU, UK. GRAAF Relevant signal includes geographic specificity checklist for Dutch, Belgian, and UK market content requirements. |
| Pricing | SEO Writing Assistant included from €119/month (Pro plan). Content Marketing Platform from €449/month. Ongoing subscription required. | ContentScore scan free at app.contentscale.site. No subscription. GRAAF implementation guidance available through my consultancy at contentscale.site/services/. |
| Best for | In-house teams needing keyword data, rank tracking, and competitor research at scale. Strong data infrastructure for established SEO teams. | Publishers who have lost traffic to AI Overviews, have content that ranks but doesn’t get cited, or need market-specific guidance for NL/BE/LU/UK audiences — and want a free diagnosis before spending anything. |
Using Semrush and GRAAF Framework together
Semrush and GRAAF solve different problems. Semrush gives you keyword volumes, competitor rank tracking, and content recommendations based on what currently ranks. GRAAF tells you why AI systems are not citing your page despite ranking — and gives you the specific five-signal checklist to fix it.
The most common pattern: a page ranks position 5–15 for a target keyword, Semrush’s Content Score shows 80+, but the page receives near-zero CTR because AI Overviews absorb the query above it. That is not a keyword density problem (what Semrush measures). It is a Genuinely Credible + Accurate signal problem (what GRAAF measures). The free ContentScore scan shows which of the five GRAAF signals your page is failing in 30 seconds.
❓ Frequently Asked Questions: GRAAF Framework
Quick Answer: A 5-signal content methodology — Genuinely Credible, Relevant, Actionable, Accurate, Fresh — that maps directly to the quality dimensions AI Overviews evaluate before citing a page.
The GRAAF Framework was created by me after analysing why AI systems cite some pages and ignore others across identical topics. The five signals emerged from my traffic recovery implementations as the consistent differentiators between cited and non-cited content. Pages scoring strongly across all five achieve strong AI citation rates when combined with CRAFT editing — versus under 10% for pages missing two or more signals. The framework is implemented through three phases: GRAAF quality foundation (Phase 1), CRAFT editing (Phase 2), technical SEO (Phase 3).
Quick Answer: E-E-A-T is conceptual guidance without implementation steps. GRAAF provides 5 specific checklists, a 100-point scoring system, and documented citation rate outcomes — all designed specifically for AI Overviews.
Google’s E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) defines what quality content looks like in principle. It does not tell publishers how to build those signals into a page, how to measure them, or how they map to AI citation specifically. GRAAF fills all three gaps. It builds on E-E-A-T concepts but adds the implementation layer that practitioners need: concrete checklists per signal, a ContentScore that tracks progress, and a 3-phase workflow that sequences the work correctly. See the comparison table in Section 4 for a full breakdown of differences.
