Every study on this page was designed and run by Blogging Titan. We collected the data ourselves, from live search results, AI answers, vendor pricing pages and our own server, and each write-up explains exactly how. Nothing here is a reader survey or a rewrite of someone else’s report.
- 42 studies and data pages
- First published June 2026
- Free to cite with a link
Jump to: How AI search chooses its sources · Can you trust the advice on page one? · What blogging tools really cost · Tests run on our own site
How AI search chooses its sources
What Google's AI Overviews, AI Mode, ChatGPT and Perplexity cite, how often that changes, and which page features make no difference.
- AI citations per answer: Conductor's US benchmarks against our blogging sample
- Which UK businesses does AI recommend? 160 Google AI Mode answers
- Can AI give UK businesses accurate marketing prices? 220 answers re-run
- The Citation Lottery, 87 days later: 65% of July's AI Overview citations are gone
- 224 of the 951 pages AI search cited carry no publication date
- Who wrote the 951 pages ChatGPT and Perplexity cited? On 195, nobody
- Reddit ranks on all 30 blogging searches; Google's AI quotes it on 18
- The AI Citation League Table: 100 blogging questions put to ChatGPT and Perplexity
- Ask Google the same question three ways and only 14% of its sources survive
- The Dateless Third: a third of what Google's AI cites has no publication date
- How old is the content Google's AI cites? 243 citations dated
- The Citation Lottery: the same 20 AI Overviews pulled three times in one hour
- AI Overviews do not reward fresh content, they punish stale content (156 pages)
- Schema markup does not predict AI Overview citations (138 pages)
- Google AI Overview and Perplexity barely cite the same sources
- The First-Person Fallacy: does "I tested" language win AI citations?
- The Rank-Citation Disconnect: 68% of AI Overview citations are outside Google's top 10
Can you trust the advice on page one?
Audits of the pages that already rank for blogging questions: who wrote them, who profits from them, and whether their numbers hold up.
- Eight AI search experts on getting cited, checked against our own data
- State of UK Blogging 2026: 174 page-one pages, 24 of them British
- We counted the statistics on 161 page-one blogging pages. Seven in ten cite nothing
- AI traffic converts 1.3x better. Or 4.4x. Or 23x. Every claim traced to source
- The Reddit Tax: Reddit is on page one for all 30 blogging searches we checked
- 369 prices in 39 page-one tool reviews: mostly accurate, rarely usable
- The Ghost Author Test: on 15 of 48 page-one guides there is nobody to find
- The Proof-of-Use Test: 1 in 5 page-one tool reviews shows no sign of use
- The Fake Refresh: 14 of 25 "Updated 2026" posts changed almost nothing
- The Stake Test: only 8 of 40 "best blogging tool" posts had no stake in the answer
What blogging tools really cost
Prices read from vendors' own pricing pages, dated, and kept as a running history.
- How the gaming blogs that rank make money, read from their own ads.txt files
- What UK business marketing really costs: the prices companies publish
- Five AI writing tools (eight models), one brief, every word scored
- A year of 16 AI SEO and blogging tools: for 13, monthly price times 12 is wrong
- Pricing pages: every plan and a dated price history for each tool we cover
- Blogging Tool Cost Index 2026: what 30 tools actually cost
Tests run on our own site
Experiments where Blogging Titan is the test subject: server logs, our own link graph and our own writing.
- What Cloudflare Pay Per Use offers a small blog, and why ours cannot see an offer
- Cloudflare let ChatGPT back in on 15 September. On our open blog, its fetches did not move
- Every AI crawler on a small blog, logged for 37 days
- We linked to our best page 25 times. The live HTML showed zero
- All 1,403 internal links on this blog mapped: one in five takes a redirect detour
- The Signs of AI Writing checklist, run on 176 of our own posts
- The orphan-page problem: an internal-linking audit of 153 posts
- The Answer-First Test run on 152 blog posts
- Click-through rate by Google position, from a real blog's Search Console
How these studies are made
The short version: we fix the question and the sample before collecting anything, record every observation with its date, and publish the counts with the method beside them. When a result goes against what we expected, we publish that too. The Teaser Rate study set out to show that tool reviews quote stale prices; the data said most do not, and the post says so.
The full process, including how we check our own figures and the mistakes those checks have caught, is on How we test. Who writes and edits the site is on our editorial policy.
How to cite this research
You can quote any figure from these studies, in an article, a deck or an AI answer, under a Creative Commons Attribution 4.0 licence. Credit “Blogging Titan” and link to the study page the figure came from. Most studies carry a cite-this-data box with the exact wording and date to use.
Found an error in our data? Email contact@bloggingtitan.com. Confirmed corrections are made on the page and listed on How we test.
For the headline figures in one place, see AI Overview and GEO statistics.