What Is The Faunalytics Lens?
The Faunalytics Lens is an AI-powered research and synthesis tool that combines the capabilities of frontier AI with a focused knowledge base of Faunalytics’ Research Library, Original Research, and all of the other resources we publish. Unlike a typical website search function — which would provide a list of posts based on your search terms — the Lens can engage in conversation on any animal advocacy topic and provide answers in a synthesized and summarized form.
Read the full launch blog: Look Through The Faunalytics Lens.
How Does It Work?
When you submit a question or start a conversation, the Lens connects to a database of all Faunalytics resources, which are scraped from our site and tagged with granular metadata. Using that as a knowledge base, the Lens then generates a synthesized response. You can ask further questions to dig up more resources, take the conversation in a different direction, and otherwise continue using it until your research is complete.
It’s important to note that, unlike when you ask a standalone LLM a question, the Lens is not retrieving our resources (or any other resources) by searching the web — this can produce unpredictable errors, incorrect citations, and hallucinated information. Instead, the Lens is retrieving information strictly from our database, which allows for highly accurate responses.
The Lens “persona” is based a system prompt that has been developed to provide the best possible responses in the tone you’ve come to expect from Faunalytics: friendly, informative, and plain-spoken. It also includes instructions on how to format lists of sources, how to reply when there is not enough data on a given topic, and more.
Technical Details
The Faunalytics Lens uses several interlocking pieces of software.
- An automatically updated, vectorized database: First, the posts and pages from Faunalytics website are scraped and vectorized into a database. Vectorizing allows each piece of information from a given resource to be tagged with metadata that ensures a very high level of accuracy in sourcing and citations. In testing, we have noticed virtually no errors in the representation of data, and the sources the Assistant provides.
- Model-Context Protocol (MCP) Server: An MCP is an open-source standard that allows AI models (LLMs such as ChatGPT, Claude, Gemini, and others) to seamlessly connect with external data sources, software, and tools. It acts as a two-way connector, enabling AI agents to access files, databases, and APIs without custom, complex integrations for each new tool. In our case, the MCP server connects to the vectorized database and uses that as its primary knowledge source.
- Large Language Model: The MCP is connected to an LLM via API calls to generate responses. There is a system prompt in place at this point in the loop that gives the LLM a persona, and instructions on how to respond. We currently use Claude Sonnet 4.5 as our model — it performed very well in testing and gives natural-sounding results that align with Faunalytics tone and style. We will continue to actively test the Lens with other models and adjust as needed.
- Website Back- and Front-End: The chat interface and design is maintained with code on Faunalytics’ website and in a private GitHub repository.
Environmental Impact
Like all online activity, the Faunalytics Lens has an environmental footprint. Currently, the Lens leverages Claude Sonnet 4.5. Its predecessor, Sonnet 4, ranked highest in eco-efficiency among large language models with a score of 0.886, combining strong reasoning capabilities with an efficient infrastructure footprint. Anthropic claims to work with cloud providers that prioritize renewable energy and carbon neutrality with the goal of maintaining net zero climate impact on an annual basis, comprehensive environmental disclosure remains limited across the AI industry. There are no reliable estimates we’re aware of that define impact on a per-query basis. Furthermore, estimating things on a per-query level can be misleading if that estimate doesn’t include what it takes to make that query possible, including vast infrastructure builds, and the impact of those. Faunalytics would welcome thorough and accurate Life Cycle Assessments of AI, but as yet such LCAs do not exist.
When contextualized within the animal advocacy movement’s broader mission, we believe this tool’s environmental cost is justified by its potential impact. Animal agriculture is one of the most resource-intensive industries globally, and even marginal improvements in advocacy effectiveness — enabled by faster, evidence-based decision-making — could yield environmental benefits that far exceed its operational footprint. A single successful campaign informed by better research access could prevent animal consumption that consumes exponentially more energy and water than the Lens itself uses annually. Furthermore, by consolidating research access and reducing the need for redundant literature searches, email inquiries, and duplicative analysis across multiple organizations, the tool may actually decrease the collective environmental impact of the animal advocacy sector’s online research activities.
There are open and valid questions around the environmental impact of new data centers on the local communities where they’re built; those answers are beyond our current abilities to estimate.
Run It Locally
Instructions on how to install and run this tool with your favorite LLM will be available soon.
Feedback
Do you have a question, concern, or bit of feedback about the Faunalytics Lens? Get in touch with Faunalytics’ Resource Director karol orzechowski: [email protected]
