Editing Your Brand DNA When the Extraction Isn't Quite Right

Mani's Brand DNA scan is 80-90% accurate for most websites on the first pass. That means 10-20% of the extracted profile might need your correction. This is normal and expected. The scan reads your website's HTML, meta tags, colors, and content. It does not read your mind. The 2-3 minutes you spend correcting the DNA produces hundreds of better ads from that point forward.

This guide walks through each DNA field, explains what the AI looks for, why it sometimes gets it wrong, and how to correct it for better output.

Step 1: Access your Brand DNA editor

Navigate to your brand settings in mani. Click "Brand DNA" in the sidebar. You will see your current DNA profile displayed as a card with six sections: name/tagline, tone, audience, palette, products, and visual style. Each section has an "Edit" button. You can edit one section at a time or walk through all six in sequence.

Step 2: Fix your brand name and tagline

What the AI extracts: Brand name comes from your og:site_name meta tag, or the first part of your page title (before the first dash or pipe). Tagline comes from your H1 heading or meta description.

Common issues: If your site title is "Acme Coffee Roasters | Premium Single-Origin Coffee", the AI might extract "Acme Coffee Roasters | Premium Single-Origin Coffee" as the brand name instead of just "Acme Coffee Roasters." If your H1 is a marketing headline ("Wake Up to Better Mornings") instead of a brand statement, the tagline might not represent your brand positioning.

How to fix: Click Edit on the name field. Type your brand name exactly as you want it to appear on ads. For tagline, enter the single sentence that best describes what your brand does. Keep it under 120 characters. Example: "Premium single-origin coffee delivered fresh to your door." This tagline will appear in some ad formats and grounds the AI's understanding of your core proposition.

Step 3: Adjust your tone

What the AI extracts: Tone is inferred from your website copy patterns. The AI looks for signal words: formal language (enterprise, compliance, governance) signals a professional tone. Casual language (hey, awesome, let's go) signals a playful tone. Technical language (API, SDK, documentation) signals a technical register.

Common issues: Single-page sites and landing pages often have mixed tones (formal in the header, casual in the FAQ). The AI picks the dominant tone, which might not match your actual voice. SaaS sites with developer documentation often get classified as "technical" even if the marketing voice is casual.

How to fix: Select your primary tone from the dropdown (formal, casual, playful, technical, irreverent, luxury). Then select a secondary tone. The primary tone controls 70% of the copy register; the secondary adds nuance. Most brands are not pure: "casual with technical precision" or "playful but trustworthy" are common combinations.

Test your tone setting by generating 3 ads after the change. Read the copy. Does it sound like something your brand would say? If not, try a different primary/secondary combination. The right tone makes the AI's copy feel invisible: readers should not notice it was AI-generated because it matches your existing voice.

Step 4: Refine your audience

What the AI extracts: Audience comes from explicit signals on your website ("built for teams", "designed for founders") and from domain-specific keywords (sprint, agile = dev teams; sustainable, eco = eco-conscious consumers). An LLM critic validates the proposed audience against your actual content.

Common issues: The most frequent audience error is "too generic." If the AI returns "Modern professionals" or "Businesses of all sizes," it found no specific audience signal on your site. This happens with minimalist homepages that focus on design over explicit positioning.

How to fix: Write your audience as specifically as you can in one sentence. Format: "[Who they are] who [what they care about]." Examples: "DTC beauty founders doing $500K-5M who need ad creative at 3x current volume." "Mid-career software engineers at Series A-C startups who evaluate dev tools." "Health-conscious parents in suburban areas who prioritize organic and non-toxic products."

The more specific your audience description, the better the AI writes ad copy. "Everyone" produces generic copy. "DTC beauty founders" produces copy that speaks directly to that person's daily reality. Specificity is the single highest-leverage DNA edit.

Step 5: Correct your color palette

What the AI extracts: Colors from CSS custom properties (--primary, --accent, etc.), meta theme-color tags, inline styles, and class names that suggest colors. The AI identifies primary, secondary, accent, text, and background colors.

Common issues: The AI sometimes grabs the wrong primary color. This happens when your CSS defines --primary as a background color but your actual brand color is the accent. Or when your site uses a dark theme and the AI picks the dark background as "primary" instead of the brand accent color.

How to fix: Click the color swatch for each role (primary, secondary, accent, background). Enter the correct hex code or use the color picker. The primary color appears most prominently in generated ads (CTA buttons, headline accents, border elements). The accent color is used for secondary elements. Background determines the ad's base tone (light vs. dark).

Tip: if you have a brand guide with specific hex codes, enter those directly. They override whatever the AI extracted. Your brand guide is the source of truth, not the AI's interpretation of your CSS.

Step 6: Clean up your product list

What the AI extracts: Products from navigation links, section headings, and page content. For e-commerce: product names from your catalog. For SaaS: feature names from your navigation and feature sections.

Common issues: CTA labels extracted as products ("Get Started," "Open App," "Learn More"). Generic page names extracted as products ("About," "Pricing"). Too many products listed (making generations unfocused). Not enough products (making generations generic).

How to fix: Remove anything that is not an actual product or feature name. Keep 3-8 items. For e-commerce, list your hero products (best sellers, newest arrivals, signature items). For SaaS, list your main features or product lines. The product list determines what the AI references in ad copy and which items appear in product-showcase visuals.

Less is more. A product list of "Issues, Projects, Roadmaps" (3 items) produces more focused ads than "Issues, Projects, Roadmaps, Initiatives, Cycles, Views, Filters, Templates, Integrations, API" (10 items). The AI tries to reference products in the copy. Fewer products = more prominent mentions of each.

Step 7: Validate with a test generation

After editing any DNA field, generate 3 test ads immediately. This is the fastest way to see whether your edit improved the output. Compare the new generation against ads generated before the edit. Specifically check: does the copy tone match your brand voice? Do the colors look right? Are the products referenced correctly? Does the audience framing feel targeted?

If the test generation is better, your edit worked. If it is worse or different in an unexpected way, you may have overcorrected. Dial back the edit (e.g., if you changed tone from "casual" to "formal" and the output is too stiff, try "professional" as a middle ground).

Your Brand DNA is not set-and-forget. Review it quarterly or whenever your brand positioning changes (new product launch, new audience segment, visual rebrand). A 5-minute DNA review every quarter keeps every generation fresh and accurate.

When the issue is not the DNA

Three signals indicate your generations are off but the Brand DNA itself is correct. First: when the same ad concept reads great on Instagram and stiff on LinkedIn. That is a platform-register issue, not a DNA issue. Use Campaign Studio to adjust copy register per platform rather than editing DNA. Second: when generations feel right for one product line but wrong for another (e.g., flagship products generate cleanly but accessories feel off-brand). That signals you need product-specific brand sub-profiles, not a global DNA edit. The Studio plan supports this. Third: when generations feel right today but wrong six months from now. That is brand drift, not DNA error; the DNA was correct when extracted but your brand has evolved. Re-scan and re-edit.

Knowing when NOT to edit DNA is as important as knowing when to. Over-editing the DNA chasing platform-specific or product-specific issues will degrade the global accuracy and require a full re-scan to recover.

What to do next

Once your Brand DNA is dialed in, the next workflows that compound on a clean DNA are the founder's morning queue for daily creative output, Campaign Studio editing for the platform-register adjustments mentioned above, and Brand Radar to feed competitive signal into your daily generations. If you are considering an upgrade, see pricing for product-specific DNA sub-profiles + team-shared DNA on the Studio and Scale tiers.