Measure whether ChatGPT and Gemini recommend your SaaS: 25 unbranded buyer questions, three numbers and a fixed weekly set

Search Console and rank trackers do not show whether AI assistants name your product when buyers ask for a recommendation, or which pages those assistants rely on.

1. Write about 25 questions the way a buyer searches before knowing any brand, with no brand names in them. 2. Ask each one on ChatGPT and Gemini with web search on, as a buyer would see it, and keep the full answer, every brand named, its order and every cited URL. 3. Give each answer to an LLM with a strict JSON shape (brand name and sentiment) to pull out the brands; take position from the order returned. 4. Score each brand on visibility share (answers that name it), citation share (answers that cite its domain, including subdomains) and average position only where it appears; the post shows the scoring code. 5. Keep the question set fixed each week, drop failed answers instead of counting them as misses, remember one answer is 2 points at this sample size, and read the trend. 6. In GA4, look for chatgpt.com as a traffic source. 7. Act on what was cited: get onto the third-party lists and threads the engines use, write a page for each question you lose, and add a comparison page against the most-named rival.

In the week of 12 October 2026 the author's domain-expiry monitor was named in 2 percent of 50 answers and cited as a source in 4 percent, while the top competitor was named in 38 percent. The most-cited source was a competitor's own site (102 citations), and a forum thread was cited recommending rivals. No results from the planned fixes are reported yet.

Vendor: the author builds the measurement tool used (CiteGaps) and also runs the product measured. One product, one week, 50 answers; AI answers vary between sessions, so a single week is a small sample. Answer collection is done in the vendor's tool and not shown; only the scoring code is. The fixes are a plan, not a tested result.

A list of real buyer questions, access to ChatGPT and Gemini with web search, and a way to save answers and sources.

Extract every brand or product recommended in this answer, in the order it appears. Return only JSON: {"brands":[{"name": string, "sentiment": "positive"|"neutral"|"negative"|"mixed"}]}

• DEV Community (WEBVOGUE Technologies) ↗ Vendor case study, published October 23, 2026

Source published: October 23, 2026. Last reviewed here: October 23, 2026. Spot a mistake? Tell us.