Celorly
Goals that don't die in a spreadsheet
Hello IH. First post here, so a proper introduction with the numbers attached, including the ones that do not look good.
I build Celorly, goal-setting software for teams of five to fifty. Solo, self funded, no investors. The premise is narrow on purpose: most small teams set goals once, then never reopen the file, and every tool that fixes that assumes a company with someone whose job is planning. A twelve-person company has no such person.
Where it actually stands after two months of publishing:
37 articles and a free course. 4,854 impressions in Google last month, up from 1,201. Four clicks. Average position 72, which is page seven. Twelve sessions on the whole site in thirty days, and I am fairly sure nine of those were me. Zero paying customers. Zero registered users from outside.
So on the metric that matters, nothing is working yet, and I would rather write that down than dress it up.
Then I opened the AI Performance report in Bing Webmaster Tools, which I did not know existed until three weeks ago, and found one page of mine had been cited 2,074 times in AI answers in August. Not shown. Cited as a source inside answers. On some queries we are a third of all citations. The same page has nine impressions in Google and sits at position 48.
The uncomfortable follow-up: those 2,074 citations produced one measurable session. And that page did not mention my product's name once, anywhere in the copy. Thousands of people were reading my sentences with nothing attached to them.
I fixed the name part this week. What I cannot work out is the rest of it.
So the question I actually came here to ask: has anyone turned AI citations into anything measurable? Not impressions, not vanity counts. A signup, a session you could attribute, a single person who said they found you that way. I have a channel that clearly reaches people and produces almost nothing, and I do not know yet whether that is a distribution problem, a patience problem, or a channel that simply does not convert.
5 Comments
That matches what I found, though the split in my data landed somewhere I did not expect. I tested the calculator idea before building one. Keyword Planner returned no search data at all for seven of the ten phrases in that family, and the seed "goal calculator" comes back with weight loss and savings, so in this topic the demand is not there. I do have one interactive page, an OKR scoring calculator. It entered at position 31, which reads well until you see the 28 impressions behind it.
What did separate my pages was format, not interactivity. My long essay on how to write SMART goals got 941 impressions and zero citations. The examples page, same topic, 24 concrete numeric examples in eight sections, got the 2,074. The model does not seem to want the paragraph it can write itself. It wants the specific items it cannot invent.
Templates I am mid-test on. Four files went live on 4 September with the brand in the filename, in the footer of every page and in a closing block, on the theory that a file travels to places my server never reaches. First read on 1 October.
You run several of these. Have you seen citation counts move with format, and does Bing's AI report line up with what you actually see?
I cannot separate those two yet. The name went into both pages on 3 September and I check on 1 October, a month rather than two weeks, because daily citations ranged from 1 to 460 in August and a fortnight cannot tell a real drop from noise. And what I check first is not conversion, it is whether the citations survive the change at all.
On low intent, I suspect that is the bigger half. "Give me an example of a smart goal" gets 24 examples inside the answer, so the task is finished rather than postponed. The comparison that moved me: my purchase-intent pages had 13 impressions last month and produced 1 click, while the other 4,841 impressions produced 3. So my working assumption is that this channel pays in name recognition rather than sessions, which makes the honest measure not conversion but whether anyone ever searches the name. In August, not one branded query in the whole export.