presenc.ai appeared in Google Search on March 16, 2026. That month it earned about 20 clicks. In October it earned 2,860, and the most recent 28 days came to 3,568. The site is more than 1,700 pages built with a programmatic pipeline: AI agents research, write and check each page, and a person reviews it before it ships. This page gives the numbers behind that growth, including the ones that are less flattering.
Monthly Google Clicks
| Month, 2026 | Clicks | Average position at month end |
|---|---|---|
| March (from the 16th) | 20 | About 37 |
| April | 15 | About 31 |
| May | 125 | About 11 |
| June | 380 | About 11 |
| July | 840 | About 10 |
| August | 1,390 | About 9 |
| September | 2,100 | 7.3 |
| October | 2,860 | 7.8 |
Clicks are rounded. In the 28 days to mid-November the site had 3,568 clicks and about 554,000 impressions at an average position of 7.9. A normal day now brings 100 to 130 clicks.
What Took Time
The first two months produced almost nothing. Average position moved from the thirties to about 11 in May, and clicks followed. Anyone judging the work at week eight would have called it a failure. Most of the growth came after month three.
Which Pages Earned the Clicks
These figures cover the top 1,000 pages by clicks over the period, which together earned 9,557 clicks.
| Section | Pages in the top 1,000 | Clicks | Share of clicks |
|---|---|---|---|
| Research (data and statistics pages) | 579 | 8,664 | 91% |
| Guides | 68 | 156 | 2% |
| Comparisons | 58 | 109 | 1% |
| Glossary | 60 | 92 | 1% |
| Everything else | 235 | 536 | 6% |
Pages that report specific, sourced data about a named thing did almost all the work: hardware comparisons, usage statistics, market share figures and trackers. Pages that explain a concept in general terms earned little.
How Concentrated the Traffic Is
| Group | Share of clicks |
|---|---|
| Top 10 pages | 40% |
| Top 50 pages | 68% |
| Top 100 pages | 81% |
- The median page in the top 1,000 earned 1 click.
- 344 of those 1,000 pages earned impressions and no clicks.
- 147 pages earned 10 clicks or more, and 37 earned 50 or more.
This is the normal shape of search traffic, and it matters for planning. You cannot tell in advance which pages will be the top ten. You publish a well-chosen set and a minority carry it.
Citations by AI Assistants
Ahrefs counted 446 AI-assistant answers citing presenc.ai pages in November 2026, across 103 different pages.
| Assistant | Answers citing our pages |
|---|---|
| ChatGPT | 156 |
| Microsoft Copilot | 142 |
| Google AI Mode | 81 |
| Perplexity | 31 |
| Gemini | 18 |
AI crawlers made 257,980 requests to the site in 30 days, according to Cloudflare, of which 114,810 were from OpenAI. One of our research pages was also cited in a Fox News opinion piece on November 9, 2026.
What We Did
- One page per entity. Each page is about one named thing and has its own facts.
- Sources for every figure. Each batch has a sourcing record that grades every claim.
- A link structure enforced by the build. Hubs, topic pages and related links, with checks that stop the site publishing if a page would be unreachable.
- Small batches. Tens of pages at a time, chosen from what Search Console showed was working.
- A person reviews every page.
The process is described in how to build programmatic SEO pages.
What Went Wrong
- Early pages contained unsupported figures. Some of our first batches included numbers we could not later trace to a source. We went back, corrected about 45 pages from sourced data, and removed one page about a product that never shipped. The claim-checking step exists because of this.
- Pages shipped with nothing linking to them. An audit in April found seven orphan pages. Later we found links that existed in our data and never appeared on the page.
- Forty-six research pages were published and never reached the sitemap. A build check now catches that.
- The general pages underperformed. Guides, comparisons and glossary entries together earned about 4% of clicks.
What These Numbers Do Not Show
- They are clicks, not customers. The pages that rank are about AI hardware, models and market statistics. Their readers are mostly researchers and practitioners, not buyers of our service. This shows the method can earn search traffic. It does not show lead generation.
- It is one site in one subject. AI is a fast-moving topic with constant new searches. A slower subject would behave differently.
- One day is excluded. On November 9, clicks jumped to 737 against a normal 100 or so, with impressions flat, spread across unrelated pages, and fell back within two days. We do not know the cause and have left it out of every figure here.
- Third-party tools show much less. Estimates from SEO tools are far below Search Console for this site, because the traffic is spread across many low-volume searches those tools do not track.
What We Would Tell Someone Starting
- Give it at least three months before judging.
- Put data on the page. Specific, sourced facts about named things outperformed general explanation by a wide margin.
- Expect a small share of pages to earn most of the traffic.
- Build the link structure and the checks before the second batch, not after the tenth.
- Check every claim. Corrections after publishing cost more than checks before.
To estimate what a set of new pages might do, see how to forecast organic traffic.
Methodology
Clicks, impressions and positions are from Google Search Console for presenc.ai. Monthly figures are rounded. Section and concentration figures are from a 12-month export of the top 1,000 pages by clicks, taken November 13, 2026, covering March 16 to November 10. AI citation counts are from Ahrefs and crawler requests from Cloudflare, both read in November 2026. Shares are our arithmetic. These are our own site's results and are not a prediction for any other site.