Flagship · Answer-engine visibility at scale
Cited 241,500 times. Clicked 2,500 times. Same platform.
ZPlatform.ai · 16 months · Apr 2025 – Aug 2026 241,500
Copilot citations in 6 months
~97:1
citations per organic Bing click, at minimum
SEO case studies · SaaS SEO
SaaS SEO Case Study: What 241,500 AI Citations and 8,666 Clicks Look Like Together
Part of the SEO case studies index, which publishes the raw Search Console and Bing exports for every property named here.
The short answer
One software property, measured over 16 months. Google sent 1.42 million impressions and 8,666 clicks at a 0.61% click rate, which by any traffic scoreboard is a failure. Over six months the same pages were cited 241,500 times by Microsoft Copilot, at least 97 citations for every organic Bing click. The acquisition channel moved and the dashboard I was watching never showed it.
The evidence behind this page
One row per case study, with the figure that makes it relevant to this question and the dashboard it was read out of. Every row links to the full study, where the export itself is on the page.
| Property | Figure | What it measures | Source |
|---|---|---|---|
| ZPlatform.ai | 8,666 | Google clicks from 1,420,120 impressions, a 0.61% click rate at average position 16.2 | Google Search Console, search type Web |
| ZPlatform.ai | 241,500 | Copilot citations in six months, across 294 cited pages and 1,640 grounding queries | Bing Webmaster Tools, AI Performance |
| ZPlatform.ai | ~97:1 | Citations per organic Bing click, counted over six months against clicks over 24, so a floor | Bing Webmaster Tools, both reports |
| ZPlatform.ai | 51.7% | Share of citations landing on commercial-intent queries rather than informational ones | Bing Webmaster Tools, grounding query export |
The SaaS SEO case study that reports pipeline it cannot connect to traffic
The standard SaaS SEO case study shows a traffic line going up and a pipeline number beside it, with nothing linking the two. No query export, no attribution window, no statement of what else was running that quarter.
You’re asked to assume the traffic caused the revenue.
I can’t publish client pipeline data, so this is the opposite trade. One property I own, every export attached, and no commercial metric claimed at all.
The result, stated as a failure first
Over the 16 months ending 13 August 2026, Google served ZPlatform.ai 1,420,120 impressions and sent 8,666 clicks. A 0.61% click rate at an average position of 16.2. Bing sent 2,496 clicks over 24 months.
That’s a bad SaaS SEO result. Sixteen months of hand-testing software, three to ten hours per tool, with the scoring method published, and the traffic didn’t arrive.
Then the Bing AI Performance report: 241,500 Copilot citations in six months, across 294 cited pages and 1,640 distinct grounding queries. One page alone has been cited 108,546 times, which is 45% of the site’s total.
Divide the citations by the organic Bing clicks and you get at least 97 citations per click. At least, because the citations are counted over six months and the clicks over 24. The real ratio is higher and I publish the floor.
Why hand-testing paid off in a channel I was not watching
The bet was ordinary and it was wrong in an interesting way.
The category is saturated with listicles generated from other listicles, so the one thing no generated competitor could copy was sitting down and using the software before ranking it. I expected that to win clicks, on the ordinary reasoning that more useful pages rank better and better rankings send traffic.
Being genuinely tested didn’t win clicks. It won the right to be quoted.
These are queries an answer engine can satisfy in place. Which free image generator is best. How much water does AI use. What is a token. A 0.61% click rate was the predictable outcome of answering questions that don’t require a visit, and once you accept that, the citation becomes the thing worth optimising for.
The four decisions that transfer to a product site
Publish the method, not just the verdict. Every tool is used for three to ten hours before it is ranked and the scoring rubric is public. That’s the part a generated competitor can’t copy, and it appears to be why a model treats the page as a source rather than as another aggregation of the same claims.
One comprehensive page per commercial question. “Best free AI image generators” is a single deep, maintained comparison page rather than twelve thin ones. It carries 108,546 citations because it is the most complete answer to a question people keep asking.
Structure for extraction. Comparison tables, explicit criteria, verdict lines, real pricing. 51.7% of the citations land on commercial-intent queries, which is where your revenue actually is, and those are the pages built like tables rather than like essays.
Write plainly enough to survive translation. The site is English-only and gets cited in five other languages, which only happens when the sentences are simple enough for a model to carry the fact into German or Turkish without losing it on the way.
The same four decisions produced the citation counts on two local properties with nothing else in common, written up as the local SEO case study.
What this changes about measurement
The practical finding isn’t a tactic.
It’s that your SaaS site can be winning in a channel your own analytics doesn’t display, and the default reaction to a 0.61% click rate is to conclude the content strategy failed and change it. If I’d acted on the traffic dashboard alone I would have rewritten the pages that were producing 241,500 citations.
So before you decide a page isn’t working, open the AI Performance report in Bing Webmaster Tools and the Generative AI features report in Search Console. Both are free and both are probably already sitting in an account you own.
You can’t tell whether your content strategy is working from a traffic graph any more, and you won’t find out by looking harder at the traffic graph.
Read the full case studies
Each one carries the dated screenshots, the position bands, the stated methodology and the surfaces that were not measured.
ZPlatform.ai · 16 months · Apr 2025 – Aug 2026 What this evidence does not show
Every study here is a property I own, in a vertical I happen to work in. That is the whole reason the exports are publishable, and it is also the limit.
- This is one property. A single site cannot isolate a variable, so everything here is a strongly evidenced pattern rather than a proven cause, and I have said so on the study itself.
- It is a review and comparison directory for software, not a seated SaaS product with a signup flow. It shares the buyer and the query set with one. It does not share the funnel.
- No commercial metrics are published: no cost per acquisition, no trial starts, no MRR, no pipeline. Those figures exist for client work I cannot publish, and inventing a proxy for them here would defeat the point of the exports.
- The citation evidence is Microsoft Copilot and Google AI features. ChatGPT, Perplexity, Claude and Gemini expose no comparable report, so presence there is not measured and is not claimed.
How every figure here was measured
Named narrowly. There is no crawler and no rank tracker in this list, because neither produced a number that appears on this page.
| Tool | How it was used | What it supplies |
|---|---|---|
| Google Search Console | Query, page, device and country exports pulled per property over one fixed 16-month window, then recomputed rather than read off the summary tiles. | Clicks, impressions, click rate, average position, the position bands, and the Generative AI features report. |
| Bing Webmaster Tools | Search Performance for the organic side, and the AI Performance report for citations. The AI report only retains six months, which is why the citation windows are shorter than the Search Console ones. | Copilot citation counts, cited-page counts, grounding queries, and the per-page citation breakdown. |
| Ahrefs | One dated pull per property, refreshed for the whole portfolio at once so no two figures on the page come from different weeks. | The portfolio metrics on the index page. No Ahrefs traffic estimate is published anywhere in this section. |
| The property dashboards themselves | Anything a reader can check on a public page of the property is taken from there rather than from an SEO tool. | Published stats pages, review counts, and the platform figures cited as verifiable. |
Questions this raises
Is 8,666 clicks in 16 months a good result for a SaaS site?
No, and the study says so. On the metric I originally chose to be judged by, the work failed. The reason it is published is that the same pages were cited 241,500 times over six months by an answer engine, which is a channel the traffic dashboard does not display. Reporting only the good half would make this a brochure.
Why would a directory tell you anything about SaaS SEO?
Because the query set is the same one a software company competes on: product comparisons, alternatives, pricing questions, category shortlists. 52% of the citations land on commercial intent. What it cannot tell you is anything about a trial funnel or activation, which is why that gap is listed above rather than glossed.
Can you replicate 241,500 citations on a product site?
I would not promise a number. What is transferable is the mechanism: first hand testing published with its method, one comprehensive page per commercial question instead of twelve thin ones, real comparison tables, and plain sentences that survive machine translation. That last one is why an English-only site gets cited in five other languages.
How is a citation different from a click?
A click is a visit. A citation is the answer engine using your page as a source in a response the user reads without visiting. Bing Webmaster Tools counts citations in its AI Performance report; Search Console reports AI feature impressions but not citation counts. Neither appears in a normal traffic report, which is how a channel this size stayed invisible.