Section 01Executive Summary
Generative Engine Optimisation moved decisively from experiment to line item this quarter. Brands that treated AI visibility as an instrumented channel — with panels, baselines and per-model dashboards — pulled measurably away from brands still optimising blind.
This report tracks a longitudinal panel built on the KernelX Benchmark Corpus methodology: the same prompt panels re-run across Q2 and Q3, holding intent mix and locales constant so that quarter-over-quarter deltas isolate real behavioural change. All figures are representative benchmark findings from this panel, expressed as ranges, not audited market totals.
Quarter highlights
- The share of commercial responses containing citations grew approximately 16–20% quarter over quarter — assistants are grounding more of their commercial answers.
- Official websites widened their lead over blogs as citation destinations; generic editorial content lost share for the second consecutive quarter.
- Structured comparison pages outperformed generic landing pages by an expanding margin in best-of and comparison intents.
- Brands with consistent entity information across their site, knowledge panels and third-party profiles achieved higher and more stable recommendation rates.
- Knowledge-graph coverage continued to expand into mid-market brands, raising the entity-resolution baseline across most industries.
- Content freshness signals strengthened: pages updated within the quarter were cited disproportionately in fast-moving categories.
Key takeaways
- GEO matured into a measurable channel this quarter; grounding and citation behaviour intensified.
Business implications
- The window for "easy" visibility gains from basic entity hygiene is beginning to close in competitive categories.
Recommendations
- If a brand has no per-model baseline yet, establishing one is now the single highest-leverage action.
Section 02Methodology Note
The full measurement methodology — prompt taxonomy, entity registry, recommendation scoring, the Visibility Score and its weights — is defined in the flagship paper We Analysed One Million AI Responses. This quarterly applies that framework as a fixed longitudinal panel: identical prompt panels, identical intent mixture, identical locales, re-sampled each quarter.
Deltas are the product here, not levels. A fixed panel makes quarter-over-quarter change interpretable; individual point estimates inherit all the limitations documented in the methodology paper.
Key takeaways
- Fixed panels isolate behavioural change from measurement noise.
Business implications
- Teams replicating this design can compare their internal deltas against these industry baselines.
Recommendations
- Never edit a trend panel mid-stream; version it and restart the baseline instead.
Section 03AI Visibility Trends
Median Visibility Scores rose modestly across the panel, but the distribution widened: leaders compounded while the long tail stagnated. The clearest single trend is the growth of grounded, citation-bearing answers.
The visual communicates a steady rise in grounded answers across the half-year, accelerating in the final two months of the quarter.
The visual communicates widening inequality of visibility: gains concentrate in brands already investing in entity and evidence infrastructure.
Visibility is compounding. Brands cited this quarter are more likely to be retrieved, recommended and cited again next quarter — consistent with the layered Entity Authority Model from the methodology paper.
Key takeaways
- Grounding is rising; visibility gains concentrate among prepared brands.
Business implications
- Late movers face a steepening curve as leaders compound authority.
Recommendations
- Prioritise becoming citable (canonical, verifiable pages) over becoming louder.
Section 04Citation & Source Trends
Within cited responses, the mix of destinations shifted toward canonical and data-led sources. Official sites extended their lead; generic blog content continued its decline; original research strengthened for the second straight quarter.
| Source class | Q2 share | Q3 share | Direction |
|---|---|---|---|
| Official brand sites | ~77–79% | ~80–82% | ▲ rising |
| Community (Reddit et al.) | ~35–38% | ~37–40% | ▲ rising |
| Encyclopedic (Wikipedia) | ~32–34% | ~32–35% | → stable |
| Review platforms | ~29–32% | ~28–31% | → stable |
| News & trade media | ~24–27% | ~23–26% | → stable |
| Independent blogs | ~19–21% | ~16–18% | ▼ declining |
| Original research & data | ~7–8% | ~9–11% | ▲ rising fast |
Table 1 — Citation-destination share among cited responses, Q2 vs Q3 (classes are not mutually exclusive).
Freshness
Recency now behaves like a genuine ranking factor in fast-moving categories: pages updated within the quarter were cited at roughly 1.5–2× the rate of otherwise comparable pages last touched a year ago. The effect was strongest in AI tooling, finance and travel; weakest in slow-cycle industrial categories.
The declining citation share of generic blogs is not a penalty on blogging — it is substitution. When a category offers verifiable, structured, current sources, assistants use them first. Editorial content earns citations only where it adds evidence, not volume.
Key takeaways
- Canonical and data-led sources gained share; generic editorial lost it; freshness matters more.
Business implications
- Content roadmaps weighted toward interchangeable posts are depreciating assets.
Recommendations
- Refresh cornerstone pages quarterly; convert top editorial themes into data-backed assets.
Section 05GEO Ranking-Factor Shifts
Re-estimating the signal correlations from the methodology paper on this quarter's panel shows a stable top tier and meaningful movement underneath it.
| Signal | Q2 strength | Q3 strength | Movement |
|---|---|---|---|
| Entity completeness & consistency | Strong | Strong | → holds #1 |
| Structured comparison content | Strong | Strong | → holds |
| Third-party citation footprint | Strong | Strong | → holds |
| Original research & data assets | Moderate | Moderate–strong | ▲ rising |
| Content freshness | Weak–moderate | Moderate | ▲ rising |
| Community presence | Moderate | Moderate | → category-dependent |
| Classic technical SEO | Moderate | Moderate | → necessary, not sufficient |
| Publishing volume alone | Weak | Weak | ▼ weakest measured signal |
Table 2 — Correlation strength of brand-side signals with recommendation share, Q2 vs Q3. Correlational, not causal.
Technical SEO vs GEO
The two disciplines are converging into a stack rather than competing: crawlability, structured data and performance determine whether AI crawlers can ingest a brand's evidence; GEO determines whether that evidence wins the answer. AI crawler activity in panel-site logs continued to grow this quarter, and brands blocking those crawlers wholesale showed measurably lower retrieval rates.
Audit robots and CDN rules for AI crawler handling deliberately, not by default. Blanket blocking trades long-term visibility for short-term content control — a reasonable choice only if made consciously.
Key takeaways
- The signal hierarchy is stabilising; research assets and freshness are the movers.
Business implications
- GEO budgets can now be planned against a reasonably stable factor model.
Recommendations
- Treat technical SEO as the ingestion layer of GEO, owned by the same roadmap.
Section 06Industry Scorecard: Winners & Losers
Aggregating Visibility Score changes across the 35 tracked industries produces a clear pattern: categories with active research cultures and dense comparison content matured fastest.
The visual communicates maturity gaps by Entity Authority layer: software categories lead on every dimension; industrial categories lag most on consensus and freshness.
| Movement | Industries | Driver |
|---|---|---|
| Biggest gainers | AI developer tools · fintech infrastructure · cybersecurity | Dense comparison content, active communities, frequent original research |
| Steady leaders | CRM · cloud platforms · e-commerce SaaS | Mature entity coverage; gains now incremental |
| Under-performers | Traditional insurance · industrial equipment · local services | Thin canonical evidence; minimal third-party consensus |
| Most volatile | Consumer AI apps · creator tools | Rapid launches outpace entity establishment; high hallucination exposure |
Table 3 — Q3 industry movement summary from the longitudinal panel.
Key takeaways
- Industry maturity diverged further; volatility concentrates where entities are youngest.
Business implications
- Category context sets realistic targets: median visibility in SaaS ≠ median in industrial.
Recommendations
- Benchmark against your industry's scorecard row, not the global median.
Section 07Model Updates, AI Overviews & Enterprise Adoption
Two model-family updates during the quarter measurably reshuffled recommendations in fast-moving categories — reinforcing the case for continuous rather than campaign-based measurement. Search-embedded AI overview surfaces continued to expand coverage of commercial queries, blending classic ranking signals with assistant-style synthesis and pulling the two disciplines closer together.
On the demand side, enterprise adoption of GEO practices accelerated: within the panel's enterprise segment, the share of brands showing evidence of deliberate GEO work (fresh comparison hubs, structured data expansion, entity-consistency cleanups) grew from roughly a fifth to a third quarter over quarter. Agencies report the same shift — per-model differences are becoming a standard client question.
Post-update volatility is a leading indicator worth monitoring: categories whose rankings reshuffle most after model updates are the categories where authority is shallowest — and where fast movers can still leapfrog incumbents.
Key takeaways
- Model updates reshuffle shallow categories; AI overviews blur search and assistants.
Business implications
- Quarterly campaigns cannot track weekly model churn; monitoring must be continuous.
Recommendations
- Alert on post-update visibility deltas the way SRE teams alert on regressions.
Section 08Q4 Predictions & the 12-Month Agenda
Q4 predictions
- Citation growth continues: grounded commercial answers exceed two-thirds of the panel by year end.
- Research arms race begins: more brands publish original data assets as the citation premium becomes widely known; early-mover advantage narrows.
- Entity tooling consolidates: knowledge-graph and structured-data hygiene becomes productised, raising the baseline and shifting differentiation up the funnel.
- Agentic evaluation appears: first measurable traces of autonomous agents consuming comparison content directly, previewing the convergence discussed in the methodology paper's Future of AI Visibility.
- Volatility persists in young categories: at least one major model update reshuffles consumer-AI rankings again.
Top strategic priorities for the next twelve months
| # | Priority | Funnel layer | Expected payoff |
|---|---|---|---|
| 1 | Establish per-model visibility baselines on a fixed prompt panel | Measurement | Every subsequent decision becomes evidence-based |
| 2 | Entity consistency cleanup across site, profiles and knowledge graphs | Identity | Raises resolution rate; prerequisite for everything above it |
| 3 | Build structured comparison and evaluation hubs | Evidence | Largest observed lever on best-of and comparison intents |
| 4 | Publish at least one original research or data asset | Consensus | 2–3× citation premium over standard editorial |
| 5 | Quarterly freshness cycle for cornerstone pages | Evidence | Compounding recency advantage in fast categories |
| 6 | Earn independent mentions (communities, trade press, expert reviews) | Consensus | The signal owned channels cannot substitute |
| 7 | Deliberate AI-crawler policy and structured-data coverage | Ingestion | Prevents silent exclusion from retrieval |
Table 4 — The twelve-month GEO agenda, ordered by dependency: measurement first, then identity, then evidence and consensus.
Key takeaways
- Q4 favours brands that industrialise: fixed panels, entity hygiene, research assets, freshness cycles.
Business implications
- GEO is transitioning from tactics to operating discipline; organisational ownership matters.
Recommendations
- Assign explicit ownership of AI visibility with a quarterly review cadence — the brands compounding fastest already have.