Methodology, stated before the numbers
Sendr.ai is a B2B SaaS company and a RankedTag client. During the engagement, its signup workflow posted an alert to a production channel for every new signup, including the attribution source string captured at signup and the country reported with it. The screenshots of 17 of those alerts are published on the RankedTag homepage, with customer names, emails and phone numbers blurred.
This page is the dataset behind those screenshots. Each row is transcribed from one alert. The source string is recorded exactly as the workflow wrote it, which is why the table below contains both "ChatGPT" and "chatgpt", and why "SEO", "search" and "AI" appear as sources even though they name no engine. Normalising them would make the table tidier and the evidence weaker.
How the sample was selected is the part a reader should weigh most heavily: these are the alerts that were captured as screenshots, not a systematic export of all signups in a defined period. That makes it a convenience sample. The client reports materially higher monthly signup volume than this sample contains, so the 17 alerts here are a subset of an unknown larger population, and nothing about proportions in the wider population can be inferred from them.
Attribution source, as recorded
| Source as recorded | Alerts | Share of this sample | Type |
|---|---|---|---|
| Gemini | 4 | 24% | Named AI assistant |
| 4 | 24% | Search or generic string | |
| ChatGPT | 3 | 18% | Named AI assistant |
| Claude | 2 | 12% | Named AI assistant |
| AI | 1 | 6% | Search or generic string |
| NotebookLM | 1 | 6% | Named AI assistant |
| search | 1 | 6% | Search or generic string |
| SEO | 1 | 6% | Search or generic string |
Read that table as counts out of 17, never as market share. With a sample this size a single alert moves a percentage by roughly six points, which is why the counts are printed beside every percentage.
Country, as reported
| Country code, as reported | Alerts |
|---|---|
| FR | 3 |
| GB | 3 |
| IN | 3 |
| US | 3 |
| CA | 1 |
| CH | 1 |
| DE | 1 |
| EE | 1 |
| GE | 1 |
The codes are reproduced exactly as the workflow recorded them rather than being mapped to country names, because a mapping is an interpretation and this is a transcription.
What this sample supports
One thing, and it is narrower than most published commentary on this subject: named AI assistants appear as attributed signup sources in production, for a real B2B SaaS company, across several countries. Not referral traffic, not impressions, not a survey of intent. Signups, with the assistant named in the record.
That matters because the usual objection to assistant visibility is that it cannot be connected to anything commercial. These 17 alerts are a small counter-example: in this engagement, the connection existed and was recorded at the point of signup rather than reconstructed afterwards.
What this sample does not support
- No rate. Without timestamps there is no period, and without a period there is no signups-per-month, no trend and no before-and-after.
- No share of total. The sample is a subset of a larger population of unknown size, so 59% of this sample is not 59% of signups.
- No generalisation beyond one company. One B2B SaaS product, one category, one engagement. Nothing here says anything reliable about ecommerce, local services or any other vertical.
- No attribution accuracy claim. The source string is whatever the signup workflow captured, and self-reported or referrer-derived attribution is imperfect in ways this dataset cannot correct.
- No causation. These signups are not evidence that the SEO and GEO work caused them, only that the sources were recorded as shown.
A stronger version of this dataset is possible and is not published: a complete export over a defined window, with timestamps, would support rates, trends and a genuine share of total. That export is not in hand, and publishing the stronger claim before the data supports it would be the exact failure this site argues against everywhere else.
Citing this dataset
Free to quote, reproduce and build on, with attribution to RankedTag and a link to this page, under a Creative Commons Attribution 4.0 licence. Suggested wording: "In a sample of 17 production signup alerts from one B2B SaaS company, 10 recorded a named AI assistant as the attribution source (RankedTag, 19 October 2026)." If you quote the 59% figure, please carry the sample size with it.
The screenshots behind every row are on the homepage signup rail, and the search-side figures for the same engagement are in the Sendr.ai case study.
Frequently asked questions
Do AI assistants actually drive signups?
In this sample, yes: of 17 production signup alerts from one B2B SaaS company, 10 recorded a named assistant as the attribution source. That is evidence it happens, not evidence of how often it happens.
Why is the sample only 17 signups?
Because 17 alerts were captured as screenshots. It is a convenience sample rather than a systematic export, and the client reports materially higher monthly signup volume, so the sample is a subset of a larger population of unknown size.
What period does the data cover?
Unknown, and that is the sample’s main limitation. The captures carry no timestamps, so no period can be stated and no rate or trend can be derived.
Can I cite this dataset?
Yes, freely, with attribution to RankedTag and a link to this page under CC BY 4.0. Please carry the sample size alongside any percentage taken from it.
How was the attribution source captured?
The signup workflow recorded a source string with each new signup and posted it to a production alert channel. The strings are published verbatim, including the ones that name no engine.
Does this prove the SEO work caused the signups?
No. It shows which sources were recorded against those signups. Causation would need a controlled comparison that a single engagement cannot provide.