Why We Chose AI-Powered Brand Descriptions
The Problem with Extracted Descriptions
When our API extracts data from a website, we get the meta description and Open Graph description. These are the texts that site owners write for search engines and social sharing.
The problem? Most meta descriptions are terrible.
They are either too short, too generic, stuffed with keywords, or missing entirely. A significant percentage of websites have no meta description at all. Others use the same description across every page.
For API consumers building CRM enrichment tools or company directories, these descriptions are not good enough. They need a clean, consistent, professional summary of what each company does.
Enter AI Enhancement
We added an AI Enhancement layer that takes all the extracted data (logos, colors, fonts, meta tags, page content) and generates a polished brand description using large language models.
The generated descriptions are:
- Concise: 2-3 sentences summarizing what the company does
- Professional: written in a neutral, informative tone
- Accurate: based on actual page content, not hallucination
- Consistent: following the same structure for every brand
Here is an example for Stripe:
Stripe is a financial infrastructure platform that enables businesses to accept payments, manage subscriptions, and handle complex financial operations through developer-friendly APIs. Founded in 2010, the company serves millions of businesses worldwide, from startups to Fortune 500 companies.
Compare that to Stripe's actual meta description, which focuses on SEO keywords rather than a clean summary.
How It Works
When you pass enhance: true in your API request, we send the extracted data through our AI pipeline:
- Context Assembly: We compile the extracted meta tags, page title, headings, and body text into a structured prompt
- Description Generation: A large language model generates a brand description based on the extracted context
- Color Validation: The AI cross-references extracted colors against the visual content to correct any extraction errors
- Confidence Scoring: The model assigns a confidence score based on how much data was available and how consistent it was
{
"ai": {
"description": "Stripe is a financial infrastructure platform...",
"confidence": 0.92,
"colors_corrected": false,
"language": "en"
}
}
Multilingual Support
One of the most requested features since launch has been multilingual descriptions. With AI Enhancement, we can generate brand descriptions in any language:
curl -X POST https://api.fetching.company/v1/analyze \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"url": "https://stripe.com", "enhance": true, "language": "nl"}'
This is particularly valuable for:
- International CRM platforms serving users in multiple languages
- Localized directories that need descriptions in the local language
- Market research that requires translated brand summaries
The AI generates native-quality descriptions, not machine translations. The model understands context and produces natural-sounding text in the target language.
Color Correction
An unexpected benefit of AI Enhancement is improved color accuracy. By analyzing the visual content of the page, the AI can detect when our CSS-based color extraction returns incorrect results.
Common correction scenarios:
- A site's CSS uses white as the background but the visual design uses a dark hero section
- Brand colors are defined in a JavaScript framework and not directly visible in CSS
- Multiple color themes exist and the AI selects the most prominent one
Cost and Credits
AI Enhancement costs 2 credits per API call instead of 1. We chose this pricing because the AI processing adds significant computational cost, but we wanted to keep it accessible.
For most use cases, the additional credit is worth it. The improved descriptions and color corrections significantly increase the quality of the data.
Try It Today
AI Enhancement is available on Pro plans and above. Upgrade your account or sign up to start using AI-powered brand intelligence.
