Running your first Brand DNA extraction takes 90 seconds. But understanding what happens during those 90 seconds will help you interpret the results and make better edits. This module walks through the extraction process step by step, explains what each extractor does, and shows you how to read the results.
Before you start: preparing your URL
The quality of your Brand DNA extraction depends on the quality of the URL you provide. Mani scrapes your homepage and up to 5 linked pages (About, Products, Pricing, Blog, Contact). The richer these pages are in content, the better the extraction.
Best URLs to use: your homepage if it has substantial text content, or your main product page if your homepage is minimal. Avoid URLs that require authentication, load content entirely via JavaScript without server-side rendering, or are behind Cloudflare's aggressive bot protection.
If your website is a single-page landing with fewer than 200 words, the extraction will produce results but they may be thin. In that case, you will supplement with manual edits (covered in the next module).
The extraction pipeline: what happens in 90 seconds
When you paste your URL and click Extract, four things happen in sequence:
Stage 1: Scraping (10-15 seconds). Mani visits your URL and follows up to 5 internal links. It extracts: all visible text content, CSS stylesheets (for color extraction), image URLs, meta tags, and structured data. The scraper uses a standard browser user agent and respects robots.txt directives.
The scraping stage is the foundation. Everything that follows depends on the quality of the scraped content. If the scraper misses content (because it is loaded via JavaScript, behind authentication, or on a subdomain), the downstream extractors will have less material to work with.
Stage 2: Parallel extraction (20-40 seconds). Six extractors run simultaneously, each processing the scraped content from a different angle. This is the core innovation in Mani's pipeline. Running extractors in parallel reduces the total time from 5+ minutes (sequential) to under 40 seconds.
The tone extractor analyzes sentence structure, vocabulary complexity, formality level, and recurring adjectives. It produces 3-5 tone descriptors that characterize your brand voice.
The audience extractor identifies who your products serve by analyzing product descriptions, pricing language, and benefit statements. It produces a specific audience description with demographics, psychographics, and pain points.
The palette extractor parses your CSS stylesheets to find declared colors: custom properties, background colors, text colors, and accent colors. It groups these into 5 roles: primary, secondary, accent, text, and background.
The product extractor identifies products by scanning navigation, product pages, and pricing sections. It captures product names, descriptions, and prices where available.
The keyword extractor identifies the 10-20 most frequently used terms that are specific to your brand and industry. It filters out common words and focuses on terminology that distinguishes your brand.
The visual style extractor analyzes image patterns, layout structures, and design conventions across your pages. It produces a description of your brand's visual approach.
Stage 3: Critic layer (5-10 seconds). A critic model reviews each extraction for specificity. Generic outputs get flagged and re-extracted with tighter constraints. If the audience extractor returns "modern professionals," the critic flags it as too generic and triggers a re-extraction with instructions to be more specific. This critic pass prevents the bland, one-size-fits-all descriptions that plague most AI tools.
Stage 4: Assembly (1-2 seconds). All six extractions merge into a single Brand DNA card and are saved to your account. The card is immediately available for generation.
Reading your results
After extraction completes, you see your Brand DNA card with six sections. Here is how to evaluate each one:
Tone: Do these adjectives describe how you WANT your ads to sound? Note that your website tone and your ad tone do not need to match. If your website is formal but you want casual ads, you will edit the tone in the next module.
Audience: Is this specific enough? "DTC founders doing $1-10M on Shopify" is specific. "Businesses" is not. The more specific the audience, the better the generated copy. If the audience is vague, this is the highest-priority field to edit.
Palette: Do these 5 colors match your brand? Click each swatch and compare to your actual brand colors. Palette extraction is usually 95%+ accurate because colors are objectively defined in CSS. If a color is wrong, it is easy to fix with a hex code.
Products: Are your key products listed? If products are missing (common for large catalogs or JavaScript-loaded product pages), you will add them manually. Products that appear in your Brand DNA get referenced in generated ads.
Keywords: Do these terms reflect your brand's vocabulary? Remove any that feel off-brand and add industry terms the scanner may have missed.
Visual style: Does this description match your photographic and design approach? This field has the most room for interpretation, so manual refinement is common.
Common first-extraction patterns
After running thousands of extractions, we see consistent patterns in first-time results:
Brands with content-rich websites (detailed product pages, an about page, blog posts) produce excellent Brand DNA on the first pass. The extraction accuracy is typically 80-90%, requiring only minor edits.
Brands with minimal websites (one-page landing, coming-soon page) produce thinner results. The tone and audience may be vague because there is not enough text to analyze. These brands benefit most from the manual editing covered in the next module.
E-commerce brands with large product catalogs often have accurate tone, palette, and audience but incomplete product lists. The scanner captures products from the homepage and main navigation but may miss products that are only accessible through category filters.
What to do next
Run your extraction now. Go to your dashboard, paste your URL, and click Extract. Watch the progress as each extractor completes. Then review your Brand DNA card using the evaluation framework above. In the next module, you will learn how to edit and refine each field for optimal generation quality.
Not ready to sign up? Try the free extraction at /tools/rate-my-brand. The free version produces the same Brand DNA report without requiring an account.