How AI Search Works

Understanding Housal's hybrid search technology

What is AI Search?

Housal uses hybrid AI search that combines two powerful technologies:

Semantic Search (70%)

Understands the meaning of your query. When you search for "affordable family home," it finds properties described as "budget-friendly 4BR house."

Keyword Search (30%)

Matches exact words in your query. Ensures important terms like "BGC" or "3BR" are always found in results.

How It Works

1

You Search

Type your query naturally: "Can foreigners buy property in the Philippines?"

2

AI Analyzes

Our AI converts your query into a mathematical representation (768-dimensional vector) that captures semantic meaning.

3

Parallel Search

Simultaneously runs semantic search (finds similar meanings) and keyword search (finds exact terms).

4

Smart Ranking

Results are ranked by combining scores: 70% semantic similarity + 30% keyword match. Items found in both searches rank highest.

5

You Get Results

Relevant FAQs, properties, and guides appear in < 500ms with accuracy scores.

Why Hybrid Search?

Understands Synonyms

"Cheap" finds "affordable," "economical," "budget-friendly"

Never Misses Exact Terms

Important keywords like "BGC" or "3BR" always match

Context-Aware

Understands "family home near schools" contextually

Learns Over Time

Tracks searches to identify popular queries and content gaps

Important Limitations

AI is Not Always 100% Accurate

While our AI is highly accurate (77-81% for queries found in both searches), it can occasionally misinterpret queries or return less relevant results.

✅ Always verify property details with sellers or agents before making decisions.

Only as Good as Our Data

AI search only finds content we've indexed. If we don't have information about something, the AI can't magically create it.

Not Legal or Financial Advice

AI search results are for informational purposes only. Consult professionals for legal, financial, or investment advice.

Our Technology

Housal's AI search is built with:

  • pgvector: PostgreSQL extension for vector similarity search
  • Ollama: Local LLM for generating 768-dimensional embeddings
  • PostgreSQL: Self-hosted database (no cloud vendor lock-in)
  • Infrastructure-agnostic: Runs anywhere, no external dependencies

💡 Cost: $0/month in API fees. All AI processing happens locally on our servers.

⚠️ AI-Powered Search

This response is generated by AI and may not be 100% accurate. Always verify property details directly with sellers or agents before making decisions.

Learn how AI search works