12 of 12 model-runs expect this · p 0.50
English Paper 1 one common paper — Advanced and Standard together
Paper 1 (unseen texts + the common-module essay) is one common paper for both English courses, so these predictions pool both cohorts' six-model panels — twelve model-runs over 66 Paper 1 question-level predictions. Whatever course you're in, this page is your Paper 1; your Paper 2 styles live on your course page.
Built by a six-model AI panel and backtested against the hidden 2025 papers — how we did it.
The near-certainties
11 of 12 model-runs expect this · p 0.58
Visual + written text comparison returns as a 5–6 mark discriminator
10 of 12 model-runs expect this · p 0.80
Structural: ONE common essay question (no per-text bank), five-question Section I, 3/4/4/4/5
Want to check our working? every Paper 1 call, with each run's own prediction (12 runs, both cohorts)
Section II essay built on the rubric's 'anomalies, paradoxes and inconsistencies' strand 12 of 12 model-runs expect this
The whole 12-run panel converges on the same call: the common-module essay finally turns to 'anomalies, paradoxes and inconsistencies in human behaviour and motivations' — the one major rubric strand never yet used for Section II after storytelling (2020, 2025), challenges (2021), emotions/form (2022), collective ideas (2023) and qualities/motivations (2024). Paradox has been rehearsed twice in Section I (2022, 2023) — NESA's habitual promotion path — and a final-syllabus-year committee closes rubric gaps. Several runs pair it with 'challenge assumptions / see the world differently' as the yoked second idea.
Each model's own prediction
- Adv · DeepSeek V4: 'Texts reveal the inconsistencies and paradoxes of human behaviour' — to what extent (p 0.65)
- Adv · Claude Fable 5: last unexamined rubric strand (p 0.45)
- Adv · Gemini 3.1 Pro: 'fundamentally shaped by inconsistencies in behaviour and motivation' (string run)
- Adv · GPT-5.6 Sol: contradictions between qualities, motivations and actions (p 0.55)
- Adv · Grok 4.6: quoted claim + 'To what extent do you agree?' (p 0.58)
- Adv · Claude Opus 5: paradox/inconsistency + challenge assumptions (p 0.35)
- Std · DeepSeek V4: anomalies/paradoxes direct question (p 0.55)
- Std · Claude Fable 5: paradox strand promoted from Section I (p 0.30)
- Std · Gemini 3.1 Pro: inconsistencies of human motivations and relationships (p 0.90)
- Std · GPT-5.6 Sol: contradictory or unexpected responses to adversity (p 0.52)
- Std · Grok 4.6: anomalies/paradoxes + see the world differently (p 0.40)
- Std · Claude Opus 5: contradictory motives — what they invite us to reconsider (p 0.30)
Marker-feedback lineage: 2022–2025 feedback: components of multi-part stems ignored; generic 'human experience' assertions; 2022–2023 Section I feedback shows students cannot define paradox
In the practice paper: Q6
Visual + written text comparison returns as a 5–6 mark discriminator 11 of 12 model-runs expect this
A photograph or illustration paired with a prose, poetry or nonfiction extract in a 'Compare how both texts represent…' question on perception, connection, belonging or place — restoring the even-year multimodal pattern (2020, 2022, 2024) after 2025 ran all-print. Eleven of twelve runs predict it (adv-deepseek alone predicts an all-print section). Balanced treatment of both texts with sustained comparative language is the marked discriminator.
Each model's own prediction
- Adv · Claude Fable 5: photo/illustration + print, even-year pattern (p 0.55)
- Adv · GPT-5.6 Sol: prose + visual, contrasting interpretations of surroundings (p 0.76)
- Adv · Grok 4.6: Q5 photograph vs prose/nonfiction on perception of place (p 0.68)
- Adv · Claude Opus 5: compare with framing/salience/vector analysis (p 0.45)
- Adv · Gemini 3.1 Pro: two-text contrasting perspectives on environment, 6 marks (string run)
- Std · DeepSeek V4: poem + photograph on individual vs collective memory (p 0.60)
- Std · Claude Fable 5: 5-mark synthesis vs the returning visual (p 0.50)
- Std · Gemini 3.1 Pro: visual + poem synthesis, 6 marks (p 0.85)
- Std · GPT-5.6 Sol: verbal + photograph on belonging/place (p 0.48)
- Std · Grok 4.6: prose + photograph on belonging in a familiar place (p 0.52)
- Std · Claude Opus 5: written extract + photograph/illustration/cover (p 0.45)
Marker-feedback lineage: Marking feedback 2019, 2020, 2022, 2024 (unbalanced treatment, absent comparative language)
In the practice paper: Q5
A Section I item naming a paradox or duality of human behaviour 9 of 12 model-runs expect this
One short-answer item explicitly names a paradox, contradiction or duality (2019 boredom, 2022 human behaviour, 2023 consumerism, 2024 discomfort/joy) — most runs place it on a nonfiction, feature-article or poem stimulus (connection vs isolation, desire vs consequence, nostalgia vs loss). The stem's concept word must be defined and engaged, not recounted around; aligns with the essay cluster above.
Each model's own prediction
- Adv · Claude Fable 5: paradox/duality item inside the five-question mix (p 0.70)
- Adv · DeepSeek V4: belonging vs alienation via free indirect discourse (p 0.50)
- Adv · GPT-5.6 Sol: poem on a paradoxical emotional experience (p 0.61)
- Adv · Grok 4.6: 4-mark nonfiction paradox after two-year absence (p 0.62)
- Adv · Claude Opus 5: feature-article paradox, define before quoting (p 0.40)
- Adv · Gemini 3.1 Pro: paradoxes of human behaviour via narrative perspective (string run)
- Std · DeepSeek V4: prose contrast revealing a paradox of belonging (p 0.65)
- Std · Claude Fable 5: 4-mark paradox/inconsistency on nonfiction (p 0.40)
- Std · Gemini 3.1 Pro: comfort-in-isolation paradox (p 0.85)
Marker-feedback lineage: 2019, 2022, 2023 feedback: students could not identify or explain the paradox
In the practice paper: Q3
Low-mark opener: explain/analyse imagery in a poem of memory, place or ritual 9 of 12 model-runs expect this
The section opens (or near-opens) with a 3–4 mark 'Explain how' / 'Analyse how' on an unseen poem — free verse, first-person persona, a domestic, seasonal or remembered scene — asking how imagery, sound or enjambment conveys a named emotion or insight (memory, awe, belonging, change). Same opening-explain pattern as 2021–2025 Q1. Accessible marks, but recount instead of technique-to-effect is flagged every single year.
Each model's own prediction
- Adv · Grok 4.6: Q1 3-mark poem, explain imagery/sound for a named emotion (p 0.82)
- Adv · DeepSeek V4: childhood-memory poem, nostalgia vs loss (p 0.60)
- Adv · GPT-5.6 Sol: poem, conflicting states via imagery/contrast/voice (p 0.61)
- Adv · Claude Fable 5: poem in the 3/4/4/4/5 mix (p 0.70)
- Adv · Gemini 3.1 Pro: imagery conveying complex emotions of a new experience (string run)
- Std · Claude Opus 5: 4–5 mark analyse on a free-verse domestic/seasonal poem (p 0.70)
- Std · Grok 4.6: poem on a family or cultural ritual as shared experience (p 0.70)
- Std · GPT-5.6 Sol: 3–4 mark named-experience item, precise identification (p 0.76)
- Std · Claude Fable 5: explain on the 3-marker, analyse on 4-markers (p 0.55)
Marker-feedback lineage: 2019–2025 feedback (every year): recount over explanation/analysis
In the practice paper: Q1
A 5–6 mark 'evaluate'/'assess' judgement capstone on prose or nonfiction 5 of 12 model-runs expect this
The Advanced panel's strongest Section I call: the highest-mark item ends the section cued by 'Evaluate' or 'Assess' on a prose-fiction or literary-nonfiction extract's crafting of narrative voice, character dynamics or the human–landscape connection — following 2021 (narrative voice), 2024 (family dynamics) and 2025 (the Macfarlane landscape item). Demands an explicit judgement supported by well-chosen evidence, not analysis alone. NOTE: this cluster is carried almost entirely by the Advanced panel — see validity_flags.
Each model's own prediction
- Adv · Claude Fable 5: 5–6 mark evaluate/assess capstone (p 0.70)
- Adv · GPT-5.6 Sol: assess/evaluate on motives vs actions (p 0.66)
- Adv · Claude Opus 5: final question, 'Evaluate'/'Assess' on nonfiction/memoir (p 0.60)
- Adv · DeepSeek V4: evaluate how rhetoric positions the reader (p 0.55)
- Adv · Gemini 3.1 Pro: evaluate narrative perspective, 5 marks (string run)
Marker-feedback lineage: 2021, 2024 ('requirements for the key word assess'), 2025 ('presenting a judgement') feedback all flagged evaluative-verb failure
In the practice paper: Q4
Prose-fiction extract: voice and characterisation of an individual navigating change or belonging 5 of 12 model-runs expect this
A 4–5 mark item on a prose-fiction (or memoir) extract asking how narrative voice, characterisation and imagery portray an individual responding to change, adversity or a moment of belonging/alienation — several runs specify interiority devices (free indirect discourse, shifting focalisation) at a family gathering or on a threshold. The discriminator is explaining the resulting insight rather than recounting events.
Each model's own prediction
- Adv · DeepSeek V4: family gathering, free indirect discourse (p 0.50)
- Adv · Claude Fable 5: prose-fiction extract in the section mix (p 0.70)
- Std · GPT-5.6 Sol: voice/characterisation/imagery under change (p 0.72)
- Std · Grok 4.6: 5-mark named-human-experience prose item (p 0.90)
- Std · Claude Fable 5: one or two prose-fiction extracts in the mix (p 0.55)
Marker-feedback lineage: 2023–2025 feedback: explanation/analysis repeatedly distinguished from description and paraphrase; 2024: narrative voice under-analysed
In the practice paper: Q2
Structural: ONE common essay question (no per-text bank), five-question Section I, 3/4/4/4/5 10 of 12 model-runs expect this
Both panels' structural consensus: Section I keeps five questions totalling 20 marks (typical split 3/4/4/4/5, highest-mark item last) over five-ish unseen texts mixing poem, prose fiction, nonfiction/memoir and (per p1-q2) a visual; Section II remains ONE 20-mark question common to all prescribed texts — the per-text 6(a)–(n) bank (2019, 2022) stays rested. Common questions ran 2020, 2021, 2023, 2024, 2025.
Each model's own prediction
- Adv · Claude Fable 5: 3–6 mark five-question shape (p 0.85); common essay (p 0.75)
- Adv · GPT-5.6 Sol: five questions ~3/4/4/4/5 (p 0.78); common essay (p 0.96)
- Adv · Grok 4.6: 3+4+4+4+5 (p 0.78); one common Q6 (p 0.80)
- Adv · Claude Opus 5: 3/4/4/4/5 highest last (p 0.45); common question (p 0.75)
- Std · Claude Fable 5: single common question, no per-text variants (p 0.60)
- Std · Grok 4.6: one unified 20-mark essay (SPEC)
- Std · Claude Opus 5: one common Q6 for all fourteen texts (SPEC); 3/4/4/4/5 or 4/5/3/4/4
- Std · GPT-5.6 Sol: five questions, 20 marks, synthesis discriminator late (SPEC)
- Std · DeepSeek V4: five questions incl. comparative + paradox (SPEC)
- Std · Gemini 3.1 Pro: returns a 6-mark final synthesis item (contrarian on the 5-mark cap)
Marker-feedback lineage: Structural pattern 2021–2025; only 2020 used four Section I questions
In the practice paper: Q6
Essay framing: quote-led and/or 'To what extent…' judgement stem 9 of 12 model-runs expect this
Whatever its concept, the essay stem demands a judgement: 'To what extent…' or 'Evaluate…' rather than bare 'Analyse…', plausibly opened by a short quoted statement to negotiate (the 2023 device). Scripts that assert agreement then describe, or ignore the degree judgement, cap in the middle bands.
Each model's own prediction
- Adv · DeepSeek V4: to-what-extent paradox stem (p 0.65)
- Adv · Grok 4.6: quoted claim + to what extent (p 0.58)
- Adv · Claude Opus 5: 'To what extent…'/'Evaluate…' with personal voice (p 0.50)
- Adv · Gemini 3.1 Pro: to-what-extent stem (string run)
- Std · Gemini 3.1 Pro: evaluate-to-what-extent variant (p 0.75)
- Std · Claude Opus 5: quoted statement + 'To what extent do you agree…' (p 0.45)
- Std · GPT-5.6 Sol: quotation-led storytelling evaluation (p 0.46)
- Std · Grok 4.6: quote-led 'to what extent' (p 0.40)
- Std · DeepSeek V4: to-what-extent challenge-assumptions variant (p 0.35)
Marker-feedback lineage: 2019, 2020, 2023 (extent not engaged); 2022–2025 (personal voice and evaluation as the stated separators)
In the practice paper: Q6
Alternative essay strand: 'challenge assumptions / see the world differently' 7 of 12 model-runs expect this
The panel's second-favourite rubric hook, frequently yoked to the paradox strand in the same predicted stem: how the prescribed text invites the responder to challenge assumptions, ignite new ideas or see the world differently — requiring named form features rather than character recount. Treat p1-q1 and this cluster as one composite risk: nearly every run puts the 2026 essay on one or both of these phrases.
Each model's own prediction
- Adv · Claude Fable 5: paradoxes OR challenge-assumptions strand (p 0.45)
- Adv · Grok 4.6: challenge assumptions fallback stem (p 0.42)
- Adv · Claude Opus 5: yoked with paradox strand (p 0.35)
- Adv · Gemini 3.1 Pro: challenge our assumptions via form and feature (string run)
- Std · Gemini 3.1 Pro: challenges assumptions about collective experience (p 0.75)
- Std · Grok 4.6: challenges assumptions / ignites new ideas (p 0.45)
- Std · DeepSeek V4: challenge assumptions + see the world differently (p 0.35)
Marker-feedback lineage: 2021–2025 feedback: form metalanguage wanted over plot; 2024–2025 spent the 'insights/understanding' learner frame
Fallback essay: storytelling connecting particular lives to collective experience 7 of 12 model-runs expect this
The lower-probability alternative most runs also carry: a final-year consolidation of the storytelling strand — how the text's storytelling connects particular lives and cultures to collective experience, merging the 2020 and 2025 framings, possibly with a stimulus statement. Worth a practice essay, not the centre of preparation.
Each model's own prediction
- Adv · GPT-5.6 Sol: storytelling links particular to collective (p 0.48)
- Adv · DeepSeek V4: storytelling illuminates complexity (p 0.35)
- Adv · Claude Fable 5: storytelling through time and cultures (p 0.25)
- Std · GPT-5.6 Sol: quotation about storytelling + particular lives (p 0.46)
- Std · Grok 4.6: individual and collective via storytelling (p 0.38)
- Std · Claude Opus 5: storytelling + form metalanguage (p 0.35)
- Std · Claude Fable 5: storytelling returns after six years (p 0.25)
Marker-feedback lineage: 2020 (personal/shared storytelling), 2025 (particular lives); rubric names 'the role of storytelling throughout time'
Watch list (5)
An Australian — plausibly First Nations — voice on place or belonging among the unseen texts (adv-fable, inside its p 0.70 section mix)
A translated poem as Section I discriminator, defeating idiom-reliance (std-gemini-3.1-pro bold call, 0.30)
A 6-mark final synthesis item breaking the recent 5-mark cap (std-gemini-3.1-pro, 0.85 — sole run against the 3/4/4/4/5 consensus)
Directive-verb escalation trend: the top-mark Section I item has used evaluate/assess in 2021, 2024, 2025 (adv-fable question_type_trends)
Section II reverts to a text-specific 6(a)–(n) bank differentiated by form (adv-opus bold, 0.22; std-opus bold — both runs' contrarian hedge)
The practice paper
One Paper 1 for everyone — we don't publish rival mocks for the same common paper.
English Paper 1 (common) practice paper
40 marks · 6 questions · Stimulus booklet included.
Every question is traceable to the consensus prediction behind it — open the web version and each question carries a “why this question” link into the evidence. All questions are original Intuition compositions in NESA style.
Intu AI
One paper isn't enough? Generate more
Intu AI builds unlimited practice questions for English in these styles, marks your working, and explains what you missed — aligned to your syllabus.
Published Aug 2026, before the exams. In November 2026 we score these predictions publicly against the real paper — per-model calibration and question-level hit rates, the same harness as the 2025 backtest. How we did it.