A 404 on a heavily linked URL was turning away GPTBot on every crawl, and robots.txt was disallowing a path that held the docs. Five minutes of work, and the largest single gain on the list.
LanderLab · SaaS, landing pages
From missing in its own category to a top cited source
LanderLab ranked on page one of Google for the terms it cared about, and appeared in almost no AI answer about landing-page builders. Six weeks later it was named in 23% of tracked answers and cited as a source across the prompts that matter.
4.6xAI visibility increase
23%of tracked answers
#4position in category
6 weeksto the result
The problem
Ranking well, and invisible where it counted
The first report took thirty seconds and was uncomfortable reading. Across 42 prompts a buyer would plausibly ask — "best landing page builder for affiliates", "landing page tools with A/B testing", "alternatives to the big builders" — LanderLab was named in a small share of answers and cited as a source in almost none.
Two competitors were in nearly every answer. Neither had more traffic. Both had a comparison page for the category and a review-site profile that assistants kept quoting.
The diagnosis was blunt: of the four places an answer can lose you, LanderLab was losing at two. GPTBot was taking a 404 on a URL half the site linked to, and the pages that did get fetched carried claims with no number in them — nothing a model could lift into an answer.
What we changed
Four changes, in order of what they should move
Two of them took an afternoon. Nothing on the list was a rewrite of the site.
Every cited rival had one for this category. LanderLab did not, so there was no page for the question to map to. One page, six prompts touched.
The opening passage said the product was "a leading landing page builder trusted by modern teams". It now says what it does, for whom, at what price. Models lift sentences that carry a fact.
The review-site listing the answers quoted was out of date and missing half the feature list. Off-page, external, and worth three prompts.
The result
The same prompts, six weeks later
Same questions, same models, same daily schedule — with the day each fix shipped pinned to the chart.
Before
- Named in a small share of tracked answers.
- Cited as a source in almost none of the 42 prompts.
- Outside the shortlist when the category came up.
- GPTBot hitting a 404 on every crawl of a key path.
- No comparison page for its own category.
After
- Named in 23% of tracked answers — a 4.6x increase.
- One of the top cited sources across key prompts.
- #4 position in the category, inside the shortlist.
- Clean crawl logs across GPTBot, ClaudeBot and PerplexityBot.
- A comparison page that is now itself cited in answers.
AI is becoming a real acquisition channel for us. With ZeroRank, our AI visibility increased 4.6x and we're seeing more users come in already knowing what we do and why they want us.
Ervin HoxhaCEO at LanderLabThe method
The same four things, in every story
Different companies, different problems — but the sequence that moved the number was the same one, and it is the one you get on day one.
- Week onePrompt set built60Re-runDailyPlatforms17
Measure first. A real prompt set, on a schedule, so a change means something.
- best AI visibility trackerDailyChatGPTGeminiPerplexityClaude
Separate the two wins. Being named and being cited need different fixes.
Watch the rivals. Share of voice on the prompts that decide the category.
- Fix shippedMar 14 · 3 prompts
Keep the proof. Before and after on the prompt that failed, with the ship date on it.