fbpx

For AI-generated results to be trusted, companies must cite their sources

Man pointing to information in a book

Retrieval Augmented Generation (RAG), a type of generative AI that pulls from multiple sources to produce something new, is increasingly expected by users to cite the sources used in the generation of the new content. 

Citing sources is increasingly important from both an ethical and user experience perspective.

If you’ve been following industry discussions about Generative AI and RAG, you’re likely already aware that the output is dependent on the inputs. The AI algorithm retrieves the information it deems relevant from the sources available to it and should be able to cite those sources and present them back to the user. There have already been news stories about AI-generated content farms being used to game the programmatic ad industry, and the history of bias in AI is well documented. When sources are not made easily accessible, it can be difficult to determine the accuracy of the information presented.

In March, Google opened up Bard to users who quickly noticed that, unlike Bing Chat, Google’s Bard was less likely to cite their sources. Just two months later, Bard was updated to include better citations. Everyone from publishers concerned that AI-generated responses will hurt their web traffic to academic institutions concerned about ethics is pushing for more (and better) citations for RAG content.

Citations by AI = Transparency & Trust

We’ve talked about how important it is for companies to take proactive steps to build trust with users. Showing the content’s sources creates more transparency. Users can see where the information came from and how it was used to generate the content. This helps to build trust and credibility.

When users can see the original information, they can verify that it was used correctly. This helps ensure accuracy, and can also be helpful in determining if the source is reputable and/or if it might be biased in any way.

Sources should be cited in a way that makes sense for the user.

Adding citations doesn’t need to feel forced or visually overwhelming. Companies should provide links or reference the citation in the content itself. One example of how this can look for intelligent search results is shared below:

Visual of an intelligent search query with a RAG response that includes a list of sources

Building trust with users is essential for any company developing or using RAG. Showing the sources used to create the content is an immediate and highly visible way to help build that trust, and establish your credibility. It signals a desire to work ethically and makes it possible for anyone to ensure that the newly generated content is accurate, credible, and contextualized. This helps build trust and credibility with users, which can drive increased engagement and sales.

Contact us to discuss how we can help any size business launch AI-powered search and chat solutions that answer customers’ questions, build trust and capture leads.


* These fields are required.

July 6, 2023
Man pointing to information in a book

Trust and the Importance of AI Citations for Retrieval Augmented Generation

If you’ve been following industry discussions about Generative AI and RAG, you’re likely already aware that the output is dependent on the inputs. The AI algorithm retrieves the information it deems relevant from the sources available to it and should be able to cite those sources and present them back to the user.
June 8, 2023
Photo of a golden retriever dog jumping into a body of water

Retrieval-Augmented Generation (RAG) is the Operations Upgrade Your Business Needs

We’re living in a time when common business pain points like staffing and training can have uncommon solutions. Technology is creating new options that could not have existed just a couple of years ago. Sometimes the technology (or combination of technologies) enables solutions that were inconceivable just a couple of months ago. RAG has the potential to make the written word experiential, indexable, and immediately available for a variety of different use cases.