I test the advice before I sell it.
Short videos say what works. I check each claim against its source, and where it can be tested, I run it on a real site with the success criteria written down before the data exists. Verdicts post whichever way they land.
Inside the Lab
- 00 · tool Free ESPN fantasy hockey draft assistant (open source) Rescores ESPN projections under your league’s rules and prints the next pick during the draft. Used live on 25 September 2026, code on GitHub, misses logged.
- 01 · 24 claims Claims, checked and tested Who said it, what they sell, the verdict and the evidence. 5 have a test on a real site, 2 running now.
- 02 · 119 clips The clip library Every short video I watched and wrote up, with the useful part in one line and a link to the original so you can judge it yourself.
- 03 · first results in The Reddit interception experiment One claim, one matched pair of pages, one live site. The public log, including the reading that changed my mind.
Running now
- Queued Does IndexNow get a new site indexed faster? Starts at launch
- Running Do three-question FAQs get a page cited by AI? Running since 19 September 2026
- Queued Does an llms.txt file change what ChatGPT recommends? Choosing the test site
- Running Does putting "reddit" in a page’s URL and title get it into AI answers? Running since 12 August 2026. First readings 19 September
- Queued Does a Google Business Profile Offer bring in paid audits? Setting it up
Where 24 claims landed
- False 8
- Overstated 4
- Testing 4
- Partly true 3
- Unproven 3
- Holds up 1
- Won’t test 1
The loudest claims fall apart at the source. The useful advice is quieter.
5 claims have a test on a real site. 1 already had a verdict from the source check, so it counts under that verdict until the test reports.
All 24 claims →Grading a claim is not grading a person. Every row links to who said it. If I got something wrong, tell me and I’ll fix it and log the change.
The log
| Date | What changed | Page |
|---|---|---|
| a number moving against me | Updated the zero-click row rather than leaving it flattering. I graded "over 80% of searches end with zero clicks" false on the strength of SparkToro’s 2024 figure of about 58.5%. Their 2026 figure is about 68%. The claim is still false, and it is a lot less false than it was. Leaving the old evidence in place would have made my correction look stronger than it is. Sources: SparkToro 2024 study, SparkToro 2026 study | SEO claims |
| Checked the last nine claims sitting in the clip library’s backlog, and two of them changed what I think about a test on this page. The first: a creator promising that publishing posts as markdown brings hundreds of AI crawler hits within hours. Measured studies find close to the opposite, including one test logging zero visits to .md files against 137 to the HTML, and an analysis across 300,000 domains finding no link between these files and AI citations. That is the same family of claim as the llms.txt test below, and it means that test now has a likely answer before it starts. I am running it anyway, because a likely answer is not a measured one and my own site is cheap to experiment on, but the expectation has moved. Sources: OtterlyAI, Markdown vs HTML test, SE Ranking, llms.txt across 300,000 domains | SEO claims | |
| where that leaves the claim | The first success criterion is met and the second is not. The tactic page did reach the top 10 for its reddit query inside six weeks, comfortably, at 4th. But it has not beaten the control page’s trajectory on the control’s own target, because the control page went to 1st. The uncomfortable part for the tactic, and the most useful thing on this page so far: the control page has no reddit in its slug, title, H1 or body, and it still ranks 8th on both reddit queries. A page that ignored the tactic entirely cracked the same SERP the tactic was supposed to unlock. That does not kill the idea, because 4th in a Reddit-dominated SERP is harder won than 1st on a quiet long-tail, and the two queries are not equally competitive. It does mean the tactic has not yet shown it does anything the ordinary page did not. | Reddit experiment |
| the reading that changed my mind | The half that matters went differently than the rank numbers suggest, and I got it wrong the first time, so here is both the mistake and the finding. I asked a grounded AI surface five Shaklee questions, three runs each, and my first pass scored the pages as absent because neither URL appears in the answer text. That was the wrong test. The pages were in the retrieval set the whole time; the model just never names them out loud. Resolving the source links properly: asked what people on Reddit say about Shaklee, the model retrieves the tactic page. Asked for the pros and cons of Shaklee products, it retrieves the control page twice over, plus the older /shaklee-reviews page. So each page wins its own question. Being retrieved and being named are two different measurements and I had been conflating them; from here this log records both, separately. | Reddit experiment |
| later the same evening | Gap closed. I took the readings rather than leaving the entry above as the last word, so this page now has five-week data instead of an apology. Ranks first, organic positions with the SERP features stripped out. On "shaklee reviews reddit" the tactic page is 4th and the control page is 8th. On "shaklee reddit" the tactic page is 5th and the control page is 8th. On the control page’s own target, "pros and cons of shaklee products", the control page is 1st and the tactic page does not appear at all. Both reddit queries still show an AI Overview, and Reddit threads now hold 7 of the top 10 organic slots on each, down from 8 at day zero, with the two slots they lost going to these two pages. | Reddit experiment |
Want your own site checked this way? Get your website audited, or book a call.