Ad narrative development: benchmark methods and results · 16 October 2026
Models develop an ad's story from the fixed creative input and brand context. Ben blind-scores coherence, commercial persuasion, emotional levers and product relevance.
Reviewed 16 October 2026 · Tests and editorial direction: Ben Nancarrow · AI-assisted writing and checks.
The concept-to-brief pack fixes the creative starting point so models can be compared on developing the narrative. Coverage differs by model and repeat count; repeated versions of this pack do not represent independent audiences or brands.
What changes between prompts
The same fixed creative input and brand context are used with plain and scaffolded-context prompts. The structured condition adds guidance for developing the output. The publisher keeps hooks, narratives and compliant briefs as separate tasks, with their own scores.
How the result is scored
The narrative score averages coherence, commercial persuasion, use of emotional levers and product relevance. This measures the written narrative. The hook and the production handover have separate task scores, so a high narrative result is not inherited by the whole workflow.
All 19 reviewed model results
Human-rated narrative quality / 10
Model
Result / setup
Scored outputs
Configuration / recorded test date
Cost / scored output
Claude Opus 4.7provisional
8.58/10Simple prompt
6
thinking_high20 August 2026
Not recorded
Claude Opus 5published
8.50/10Scaffolded prompt
2
thinking_max21 August 2026
Not recorded
Claude Opus 4.8provisional
8.17/10Simple prompt
6
thinking_high20 August 2026
Not recorded
Claude Fable 5published
8.00/10Simple prompt
2
thinking_max21 August 2026
Not recorded
Claude Fable 5.1published
8.00/10Scaffolded prompt
2
thinking_max5 September 2026
Not recorded
GPT-5.5provisional
8.00/10Simple prompt
6
reasoning_high21 August 2026
Not recorded
GPT-5.6 Solpublished
8.00/10Scaffolded prompt
2
xhigh21 August 2026
Not recorded
Gemini 3.1 Pro Previewprovisional
7.83/10Simple prompt
6
reasoning_high21 August 2026
Not recorded
Claude Opus 5.5published
7.75/10Simple prompt
2
thinking_max23 September 2026
Not recorded
Gemini 3.1 Flash-Liteprovisional
7.75/10Scaffolded prompt
6
reasoning_high20 August 2026
Not recorded
Gemini 3.7 Flashpublished
7.75/10Simple prompt
2
reasoning_high21 August 2026
Not recorded
Gemini 3.8 Flashpublished
7.75/10Simple prompt
2
reasoning_high5 September 2026
Not recorded
Google Gemini 4 Argonpublished
7.75/10Simple prompt
2
high14 October 2026
Not recorded
GPT-5.4 Miniprovisional
7.75/10Simple prompt
6
reasoning_high21 August 2026
Not recorded
GPT-6 Astrapublished
7.50/10Scaffolded prompt
2
xhigh5 September 2026
Not recorded
GPT-6 Solpublished
7.50/10Simple prompt
2
xhigh23 September 2026
Not recorded
Meta Muse Spark 1.3published
7.50/10Scaffolded prompt
2
max8 September 2026
Not recorded
Gemini 3.5 Flashpublished
6.86/10Scaffolded prompt
14
Not recorded25 May 2026
Not recorded
Meta Muse Wonderpublished
6.50/10Scaffolded prompt
2
max14 October 2026
Not recorded
Recorded output counts can include repeats and both prompt conditions. Configurations and cost bases are retained; joint scores remain joint.
This record preserves the evidence and interpretation reviewed together. The live explorer can advance independently. Public results include provisional evidence; the methods above define what the scores measure.