Defining Brand Guidelines from Web Presence
A brand's web presence communicates its identity — in the language it chooses, the tensions it navigates, and the things it conspicuously never says.
This intent extracts that into an internal guideline your communications and marketing team can act on: identity, behavioral directives, tone, and voice — argued, not asserted.
It is a sharp diagnostic tool: It lets you see how outsiders actually perceive the brand and reveals the gap between self-image and external perception.
The simple (and naive) prompt
"Analyze this website and create brand guidelines based on what you see"
What could possibly go wrong? — Ways you'll recognize.
- You run the same prompt twice and get two different brand personalities — because nothing in the prompt constrains which signals the model picks up or ignores.
- The output sounds plausible but you can't tell what's actually grounded in the website of whether the model invented something. There's no way to trace a claim back to a source.
- No validation ever runs. Directives like "be authentic" ship without anyone checking whether they connect to an actual identity claim — or mean anything at all.
Sharp Definition
High value, reliable results come from being precise and explicit about what you need. The sharp definition below is ready to use — customize any value, then copy the definition, it hardly needs any prompt engineering.
Context
| Directive | Declaration |
|---|---|
| audience: | |
| purpose: | |
| purpose: | |
| background: |
Task
| Directive | Declaration |
|---|---|
| action: | |
| step: | |
| constraint: | |
| guidance: |
Input
| Directive | Declaration |
|---|---|
| source: |
Output
| Directive | Declaration |
|---|---|
| deliverable: | |
| tone: | |
| voice: | |
| format: | |
| headline: | |
| section: | |
| section: | |
| example: | |
| section: | |
| example: | |
| section: | |
| section: | |
| constraint: | |
| guidance: | |
| headline: | |
| section: | |
| section: | |
| section: | |
| section: | |
| section: | |
| validation: | |
| constraint: | |
| rule: | |
| bad example: | |
| constraint: |
Edit any value to fit your intentions, then copy the complete prompt.
Stay tuned for the full SharpIntent experience — coming soon.
Failure Modes
| Mode | Effect | Cause |
|---|---|---|
| Outsider narration | Guidelines describe the brand from the outside ("the website uses…") instead of declaring from the inside | Missing output constraint to write as if the source was not the source |
| Hallucinated values | Brand values presented as fact without grounding in published content | No guidance to distinguish observed signals from model inference |
| Slogan stacking | Directives without reasoning chain ("be authentic" with no argument) | Missing identity guidance to argue claims rather than state conclusions |
| Floating directives | Behavioral directives disconnected from identity claims | No validation rule linking behavior to belief ("Because we believe X, we always/never Y") |
| Theme-as-brand | CMS framework defaults documented as deliberate brand decisions | No distinction between intentional choices and template artifacts |
| Template output | Generic guideline with company name swapped in | No section structure, tone, or voice requirements in output spec |
| Homepage tunnel vision | Only homepage analysed, sub-brand and product pages ignored | No step to follow top-level links beyond the landing page |
Customization Guide
| What to change | Why | Where |
|---|---|---|
| Narrow the input | The default scrapes broadly. If you know where your brand voice lives, pointing the analysis there avoids noise from template pages, job postings, or legacy content that doesn’t reflect the current brand. | Change the 15-page cap and add specific URL paths to Input → source (e.g. “https://example.com — focus on /about, /team, /products, skip /blog and /careers”) |
| Add a third headline | The default structure covers identity and behavior. Brands with strong design systems need a section that connects language decisions to visual ones — otherwise the guidelines live in a silo the design team ignores. | Add a new headline entry under Output, after Behavioral Directives — e.g. “Visual-Verbal Bridge” (how verbal identity connects to layout, typography, and imagery choices) |
| Replace a section | The default assumes one consistent role. Multi-channel brands need the guidelines to acknowledge that tone shifts by surface — and define the boundaries of that shift. | Swap “What role we play in their life” for “How we show up differently across channels” if your brand varies significantly between contexts (retail vs. digital, social vs. editorial) |
| Remove the examples | The Arc’teryx-style examples set the quality bar, but they also bias the output. Removing them (or replacing with one from your own sector) lets the model find a voice that fits your brand, not mimic the sample. | Delete the two example rows under Identity if your industry is far from outdoor/technical — they anchor the model toward that register |
| Change the audience | The audience controls the level of abstraction. Internal strategists need reasoning chains. An agency team needs concrete do/don’t lists. A product team needs terminology rules and tone-per-surface guidance. Same brand, different document. | Edit Context → audience from “People defining brand and identity” to your actual readers — e.g. “Agency creative team onboarding to the account” or “Product and engineering leads integrating brand into UI copy” |
| Shift to visual design focus | The default focuses on verbal identity. If your team’s primary gap is visual consistency — how pages are built, what CMS components mean, how imagery reinforces tone — redirect the behavioral section toward design decisions grounded in the same identity claims. | Replace Behavioral Directives sections with design-oriented ones: “Layout principles and spatial hierarchy,” “Typography and type pairing rules,” “CMS component usage,” “Image selection criteria and photography direction” |