AI Redirect Mapper

Who is this for?

This tool is a game-changer for SEO specialists, web developers, digital marketers, and anyone tackling a website migration. It takes the guesswork and manual grind out of mapping thousands of URLs, saving you time and preventing costly errors.

How does it work?

1. Upload Your Lists: Drop in your old and new URLs from a CSV or XML sitemap. The tool can even use extra data like page titles to boost accuracy.
2. Generate Matches: Our AI gets to work, analyzing your lists to find the most likely redirect matches based on URL structure and content context.
3. Review & Tweak: Go through the AI's suggestions. You have full control to accept, reject, or manually edit any match.
4. Export Your Rules: When you're happy, export a ready-to-use redirects.csv.

The Tech Behind the Magic

This isn't just a simple loop. Here's a breakdown of the technical pipeline that ensures high-speed, cost-effective, and accurate URL mapping:

1. Client-Side Batching & Server-Side Indexing: Your 'Old URLs' list is batched client-side. For the first batch, the server tokenizes and builds a TF-IDF weighted inverted index of your entire 'New URLs' list. This index, a key-value map of keywords to URL data, is cached in memory for the duration of the job, making subsequent lookups for other batches near-instantaneous.
2. Hybrid Candidate Retrieval: For each old URL, we perform a hybrid search. A lexical search using the inverted index retrieves candidates based on keyword overlap, weighted by your primary factor (context vs. URL). Simultaneously, a structural analysis scores candidates based on positional path segment similarity, applying the exponential weights you've configured. The combined score from both methods produces a ranked list of the top 50 candidates.
3. Parallelized, Guided AI Inference: The 50 candidates are passed to a highly-specialized prompt. Your custom instructions are injected directly into this prompt, allowing you to guide the model's logic. Each old URL and its candidate set is processed as an independent, parallel request to the AI model, ensuring robustness and high throughput. An error in one request does not affect the others.
4. Cost & Speed Optimization: This multi-stage pipeline—from broad, cheap lexical filtering to narrow, expensive semantic analysis—is the key. By aggressively pre-filtering candidates on the server, we minimize the number of tokens and the complexity of the task sent to the LLM, resulting in faster and more affordable results than a brute-force approach.

The AI Prompt

The final and most critical step is the prompt sent to the AI. It's not just a simple question; it's a highly specialized set of instructions that turns Gemini into a URL mapping expert. The prompt commands the AI to prioritize matches based on semantic intent (e.g., product pages must map to product pages), then contextual similarity, and finally URL structure. It also enforces a strict JSON output format and includes your custom instructions, ensuring reliable and predictable results.

A Word From The Creator

Hi, I'm Søren Riisager. As an SEO expert, I built this tool to solve the exact challenges I've faced for years with technical SEO and website migrations. My goal was to create a solution that is not only effective and reliable but also genuinely saves you time.

I also created other innovative tools you might find useful, such as AskedOnce.com, Impersonal.io, and bc.tulabot.com.

Community Impact

To date, this tool has successfully matched URLs for users, saving countless hours of manual work.