Winston Francois says a client's pipeline fell 18 percent while 12 new page-one rankings stayed on informational searches

A dashboard that reports 12 new page-one Google rankings is the kind of result a founder shows the board, which is why one founder kept doing it for two quarters while his sales pipeline fell 18 percent. Jason Shafton, founder of Winston Francois, wrote in Search Engine Journal on 30 September 2026 about that client, and he describes five measures he now runs weekly to catch the gap before a full sales cycle has passed.

The disconnect in that client's data, by Shafton's account, was that every one of the 12 new rankings sat on an informational query rather than a vendor-comparison one. His team then ran the five buyer questions a prospect actually asks across ChatGPT, Perplexity, Gemini and Copilot, 20 answers in total, and the company was named in three.

The five measures Shafton runs

The first is brand presence on buyer prompts: a fixed list of five prompts across four assistants, scored every week as the share of 20 answers that name the company. Shafton says this is "the first measure to improve in our engagements, ahead of pipeline." The second is answer accuracy: for each mention, is the product right, the buyer right, the price tier right, the competitor positioning right. A Series A client appeared in 11 of 20 answers with only 4 accurate, because the assistants described it as enterprise software when it sold to businesses with fewer than 200 employees. After six weeks of correcting the public pages the assistants were drawing from, 10 of the 11 mentions were accurate, and he says "demo-to-opportunity conversion recovered the following quarter."

The third is conversion from AI referrals, segmented in Google Analytics 4 by source for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. Across his clients, Shafton reports those sessions have converted "at three to five times the rate of organic sessions." The fourth is branded search, impressions and clicks in Search Console, which he says tends to move four to eight weeks after brand presence. The fifth is what buyers say brought them in: adding "ChatGPT or another AI assistant" to inbound forms, logged in HubSpot or Salesforce. One client added that field in April, and by July 9 of 54 new opportunities had selected it. Those nine deals, Shafton writes, "closed 30% faster than the rest of the inbound group."

The timeline he gives readers

Brand presence and accuracy move first, in weeks two to six. AI referral conversions begin to show in GA4 in weeks four to eight. Branded search gains lag behind, usually weeks eight to twelve. Self-reported attribution needs at least one full sales cycle before the data is readable. The piece says to spend 40 minutes a week scoring the 20 answers in a plain spreadsheet, 15 minutes once to segment GA4, five minutes to add the form field, and leave the spreadsheet alone for eight weeks before building anything prettier.

A team that still reports Google rankings as the proof of a working funnel will keep seeing a working funnel for a quarter after sales have slowed, because the queries with most of the search volume are different queries from the ones an AI assistant answers to a buyer. Shafton's five numbers turn that gap into an early warning: a brand presence rate of 3 of 20 can be read in a spreadsheet without building a dashboard.

Source

• The AI Search Metrics I Use To Track Client Pipelines, Jason Shafton for Search Engine Journal, 30 September 2026