When a prompt mixes a brief, a brand guide, a competitor ad and your instructions, the model can confuse one for another. Wrap each piece in its own labelled tag and the confusion disappears.
Here is our brief and the brand guide and a competitor ad, write three concepts: [three long pasted blocks, one after another]
<brief>…</brief> <brand_guide>…</brand_guide> <competitor_ad>…</competitor_ad> — then: Using only the brief and the brand guide, write three concepts that could not be mistaken for the competitor ad.
The prompt
Copy-ready prompt
<brief> [PASTE THE BRIEF] </brief> <brand_guide> [PASTE THE RELEVANT PART OF THE BRAND GUIDE] </brand_guide> <reference> [PASTE A COMPETITOR AD OR A PAST CAMPAIGN] </reference> Using the <brief> and the <brand_guide>, write [NUMBER] campaign concepts. Each concept must be clearly different from the <reference>. For each: a one-line idea, a headline, and one sentence on why it fits the brief.
Why it works
Anthropic recommends XML tags because they help the model “parse complex prompts unambiguously” when instructions, context, examples and inputs are mixed. OpenAI’s guide suggests Markdown and XML for the same reason, and Google’s Gemini guide recommends consistent XML tags or Markdown delimiters.
Use it for
Any prompt with more than one pasted document. The tag names are yours to choose — just keep them consistent and refer to them by name in the instruction.
Sources
- Anthropic — Claude prompting best practices — platform.claude.com (accessed 28 Sep 2026)
- OpenAI — Prompt engineering guide — developers.openai.com (accessed 28 Sep 2026)
- Google — Prompt design strategies for the Gemini API (updated 17 Sep 2026) — ai.google.dev (accessed 28 Sep 2026)
More tricks and templates in the Prompt Lab. Want this applied to your brand’s brief? Start a project.