The Search Experiment Lab is where claims about search and AI discovery get tested. Each experiment states what I expected, what I did, the sample, the dates, what I found, what it can’t show, and gives you the data. Where I can, the plan is published before the data is collected.
Ten questions and where each stands
| Question | Status | Read |
|---|---|---|
| Does publishing more content increase AI citations? | Not tested here | None yet |
| Do backlinks influence AI recommendations? | Evidence reviewed. No test of my own | What 16 studies measured |
| Do brand mentions influence AI search visibility? | Not tested here | None yet |
| Does structured data improve discoverability? | Not tested here. One related test | What text machines read on a page |
| What happens when websites block AI crawlers? | Who blocks is measured monthly. What crawlers then do is a registered plan | The news series, the crawler log plan |
| Which publishers are cited most often? | Registered plan | Link, mention or nothing |
| Does freshness affect what AI answers say? | Tested twice | Undated figures, Bank Rate catch-up |
| How often do AI answers change? | Tested | How many prompt runs, personas |
| Do visitors from AI search behave differently? | Not tested. This site only began counting visits on 14 November 2026 | None yet |
| Does brand authority influence AI recommendations? | Not tested. AI Visibility 100 will record who gets named, which can’t show why | None yet |
“Not tested here” means exactly that. I won’t write up a result for a test I haven’t run.
Registered plans
- When AI engines repeat my findings, do they get them right and say where they came from?
- Does asking “when is that figure from?” fix an undated AI answer?
- Which UK universities block AI crawlers, month by month?
- Does an AI crawler stop when my robots.txt asks it to?
- Do the sources AI engines cite contain the figure?
- Do personas change AI brand picks outside accounting software?
- Undated current figures, every quarter
- How fast do AI answers catch up with a Bank Rate decision?
- UK news robots.txt, month by month
- Established newsbrands and lower-credibility sites
- Link, mention or nothing
- The Telegraph and the Daily Mail in AI answers
- Do news sites that block AI training still appear in AI answers?
- Do news sites enforce their robots.txt?
What every experiment includes
- Hypothesis: what I expected, written first.
- Method: enough detail to repeat it.
- Sample and dates.
- Findings, as counts with denominators.
- Limits.
- The data file.
- What I will not claim from it.
More on how studies are run on Method. The recurring research is on the UK AI Search Observatory.
Update history
- 14 November 2026: page created.