Chief was still a new product. We didn’t have name recognition or a strong, established audience. We knew AI Search was an opportunity space, but we didn’t know how to show up in the right places. This project was my attempt to figure that out.

The Goal: More Visibility on Bottom-of-Funnel AI Search

AI assistants like ChatGPT, Perplexity, and Google’s AI Overviews are increasingly where B2B buyers start their product research. If a brand doesn’t show up in the answers, they don’t enter the conversation. My goal was to restructure and create content aimed squarely at citations in bottom-of-funnel queries, the ones closest to an actual buying decision. I scoped a one-month test to find out if it would work fast enough to justify doing at scale.

The Plan: A Focused Content Sprint

I focused on six bottom-of-funnel queries tied directly to what makes Chief different: pipeline visibility, forecasting accuracy, customer success intelligence, and competitive displacement. For each one, I checked whether a page already existed that could reasonably answer it. If it did, I updated it. If it didn’t, I built new content from scratch.

The Process: End-to-End Content

This was a solo effort. Here’s roughly how it broke down:

• Research: For the more niche BoFu topics, I found the richest source material in forum threads and community discussions where people were working through the exact problem in their own words. I’d pull those conversations and use them as source material, both for the language buyers actually use and for angles existing content wasn’t covering. For competitive comparisons specifically, I went further, running sentiment analysis on Reddit and RevOps forum discussions and comparing it to G2 ratings. I found significant gaps between public rating and actual sentiment. One tool had 4.4 stars on G2 while community sentiment on it was the most polarized of any competitor in the set.
  
• Structure: Every piece followed a consistent pattern aimed at being citable, not just readable: a direct answer near the top, headers that mirrored how the query was actually phrased, and FAQ-formatted sections for the sub-questions inside the topic.
  
• Production: The brief and outline were the heaviest lifts. I figured out what needed to go into a piece based on my research and the questions users cared about. I used Claude to draft from that brief and outline, then edited heavily before anything published. I built charts in Canva and used Claude to create HTML tables.
  
• Distribution: After publishing to our site, I repurposed each piece for social content, relevant discussions on online communities, and guest post pitches to build strong third-party signals that most AI search engines weight more heavily than brand-owned content.

The Results: More Citations on Target Searches

The six BoFu queries I specifically targeted went from a combined 7 citations to 54 in the one-month window, a real signal that the approach itself works.

Important Note: Overall AI search queries barely moved in aggregate (1,011 citations to 1,076), and within that set individual queries swung hard in both directions. One term that had 326 citations dropped to zero. Others gained citations with no changes on my end at all. This illustrates just how important it is to optimize for and track the right citations. If I just tracked citations overall, I wouldn’t have results aligned to business goals.

Learnings: Figure Search Out As You Go

Targeting the queries closest to the buying decision moved the needle. When I focused specifically on BoFu terms tied to what actually makes Chief different, citations grew meaningfully in a short window. Broad definitional content didn’t show the same consistent movement.

AI citations are volatile in ways SEO rankings usually aren’t. LLMs are probabilistic engines; they don’t follow a strict input → algorithm → output formula. Citations are constantly in flux, swinging widely month to month. This is why it’s so critical to optimize for a range of relevant queries, and continue to update content over time.

I haven’t tested a key tracking assumption yet. Bing Webmaster Tools was the best citation data available for a one-month test, but it only reflects Copilot, not ChatGPT, Perplexity, or Google’s AI Overviews. I treated it as a reasonable stand-in for AI citation behavior generally. I don’t think that assumption will stand up to more rigor and scrutiny.

Treating AI Search as a monolith is a mistake. “Optimizing for AI Search” is like trying to “Win on Social.” It’s too vague a goal to be useful. In my next experiment, I want to identify which engines my audience is most likely to use, learn their quirks, and focus on producing and distributing content for those specific engines.

This site is already running that next experiment. I’m excited to see the results.

The Work: A Product-Centric Blog Post

See the content in its natural habitat: How to Choose the Best Sales Pipeline Management Software