By Ayan Chakraborty, Founder, SBA Innovation Pvt Ltd, Kolkata. Last updated . Reviewed by Ayan Chakraborty.
Why do AI assistants name only one to three lenders?
An assistant is built to answer, not to list everything. It retrieves a handful of pages, picks passages it can quote without hedging and writes a compact reply, so only the lenders that appear in those passages are named. A lender is left out when no retrieved page describes it clearly, when its facts sit in images or scripts the crawler cannot read, or when other sources describe competitors more consistently.
Because answers vary by run, the same prompt can name different lenders on different days. That is why we measure a prompt over repeated runs and count a lender as cited only when it is linked in at least two of three. The shortlist is not fixed, but it rewards the same things every time: reachable pages, direct answers with numbers, a consistent entity and independent mentions that back up your own claims.
What decides which lenders an assistant names?
Four factors do most of the work: access, extractability, consistency and corroboration. Access means the crawler can fetch the page. Extractability means the answer is a short, self-contained passage in plain HTML. Consistency means your name and facts match across your site and the web. Corroboration means independent sources mention you.
You control the first three directly and the fourth indirectly. Start with access and extractability, because they are fastest to fix and show results first, usually on Perplexity and on niche prompts. Corroboration takes months and depends on reviews, directories and coverage. Do not chase every prompt: for generic comparison prompts, aggregators with strong link profiles often win, and the effort is better spent on prompts where your own facts are the best answer.
| Factor | What it means | Lever |
|---|---|---|
| Access | The crawler can fetch your page | robots.txt and CDN rules allow search bots |
| Extractability | A short, self-contained answer in HTML | Direct answers with numbers, one question per section |
| Consistency | Same name and facts everywhere | One entity name, aligned directories |
| Corroboration | Independent sources mention you | Reviews, directories, original data |
From our client work: where we expect to lose
In the consumer lending engagement, the lender was cited with a link on 64 of 300 prompts after eight months. Two of the six head-to-head comparison prompts still went to aggregators, and we do not expect to win them.
Method: Engagement 031: median of three runs per assistant across five assistants, with a fixed prompt set and competitor list.
What did not work: Treating every prompt as winnable; the comparison prompts dominated by aggregators were not worth the content budget.
Read case study 031 Prompt set, baseline, assumptions and footnotes.