About
One person, an argument, and a stated conflict of interest.
Snippet Tester is a one-person practice. You are talking to the person who does the work, which is the only real advantage a solo operator has and I would rather lead with it than pretend to be a team.
Why the evidence-first angle
This domain used to host a product. It was a tool for A/B testing search snippets, and it got to a working beta before I killed it — because Google Search Console already reported the same numbers and my tool added a scheduler and a login screen. It solved nothing that was not already solved for free.
Making that call taught me the thing this practice is built on. The search industry is full of products and services that are technically real and practically pointless, and the only way to tell them apart is to look for evidence and be willing to accept the answer. I applied that to my own product and it failed. So the evidence base on this site is not a marketing device bolted on afterwards — it is the same test, run on everything else in the category. The first teardown I published is of the old site, for the same reason.
What I actually believe about AI search
Compressed to four sentences, so you can disagree with it before contacting me.
| Position | Confidence | Basis |
|---|---|---|
| AI chatbots are not meaningfully taking your traffic yet — all of them combined send about 0.29% of search referrals, against Google's 87.6%. | High | Referral share data, July 2026 |
| Google's own AI Overviews are taking your clicks, measurably, and that is the actual problem. | High | Ahrefs, Seer via Search Engine Land — five studies, -15.5% to -61% |
| The fixes with evidence behind them are structural and editorial: answer first, facts in tables, figures cited, headings ordered. | Moderate | Correlational: +22% statistics, +37% quotations, 2.8x heading order |
| Much of what is sold as AEO — llms.txt files being the clearest case — has no measurable effect. | High | 97% zero requests across 137,000 domains |
All four are argued with sources on the evidence page. If new research contradicts one, it changes there and the date at the top moves.
The conflict of interest
I sell the services described on this site, which means I benefit from you believing AI search matters. You should read everything here with that in mind. The specific way I try not to abuse it: the free checker will tell you when your page is fine, the services page lists who should not hire me, and the evidence page publishes findings against my own category. None of that eliminates the bias. It just makes it checkable.
How this site is built
Static HTML generated by Astro, hosted on a CDN, no web fonts, no analytics scripts, no cookie banner, no consent flow, nothing to consent to. The free checker runs as a single serverless function that fetches your page and its robots.txt and scores eight signals with regular expressions — no model, no API call, no per-run cost. That is why it is free and why it always returns the same score for the same page.
It is also the argument. If someone wants to sell you page performance, the first evidence should be their own.