6,550
outbound links checked
11,676 found; 5,126 outside the time budget
Research record · June 11, 2026
What we checked, what we counted, and where the data stops.
This is the methodology and limitation record for one bounded scan of 50 affiliate publisher sites. It documents observable link behaviour—not market-wide prevalence, customer outcomes, or revenue impact.
One observed sampleChecked-link denominator onlyNo revenue estimates
5.8%
visibly broken
380 non-blocked 4xx, 5xx, or timeout observations
9.1%
attribution failures
597 stripped-parameter or homepage-redirect observations
34/50
sites with substantive coverage
The other 16 returned fewer than three public pages and limit site-level conclusions
All rates above use the 6,550 links actually checked within budget. The 5,126 skipped links are not silently treated as healthy.
Collection method
A repeatable crawl, with the shortfalls left visible
The process was deliberately bounded. Partial coverage is part of the result, not a reason to convert missing observations into zeros.
1. Build a fixed research panel
The panel contained 50 established affiliate publisher sites across several niches. It was curated, not randomly sampled, so it does not represent every publisher or the whole market.
2. Discover a bounded set of pages
The crawler checked sitemap entries first, then used a same-site crawl when needed. It respected robots.txt and capped each site at 20 public pages.
3. Enforce the same time limits
Each site received a 100-second soft budget and a 300-second safety stop. 23 sites returned partial results; skipped links never entered a rate denominator.
4. Follow each checked link
The checker started with HEAD, retried suspicious responses with GET, allowed up to 5 redirect hops, and used a 10-second request timeout.
5. Keep blocks separate from breaks
308 responses still returned 403, 405, or 429 after the fallback. They were labelled unverifiable by the automated check and excluded from the visibly-broken count.
Counting rules
Six labels, three reporting groups
A status-code failure and a tracking failure answer different questions. They are shown separately and never converted into a financial claim.
| Observation | Count | Checked links | Reporting group |
|---|---|---|---|
| Non-blocked 4xx | 171 | 2.6% | Visible break |
| 5xx server response | 82 | 1.3% | Visible break |
| Timeout | 127 | 1.9% | Visible break |
| Tracking parameter stripped | 569 | 8.7% | Attribution failure |
| Deep link redirected to homepage | 28 | 0.4% | Attribution failure |
| Bot-blocked or rate-limited | 308 | 4.7% | Reported separately |
The central observed pattern
569 of the 597 attribution failures were tracking parameters that disappeared by the final URL. Those links can still return a successful page response, which is why status-only checking misses the category.
Read before citing
What this study cannot establish
The honest use of a bounded study is narrow. These findings describe the links observed on this panel, on this date, under these crawl limits.
Not a representative sample
The 50 sites were curated. The percentages are not a universal rate for publishers, niches, or the web.
Not complete site inventories
23 sites exhausted the soft budget, and only 34 returned at least three pages. Results are bounded observations, not full audits.
Not proof of revenue impact
The crawler did not have commission rates, traffic, conversion data, account access, or customer outcomes. No dollar inference is supported.
Not permanently current
Links, redirects, servers, and bot rules change. The record describes June 11, 2026 and should not be presented as a live measurement.
Apply the same evidence standard
Check a site without pretending the crawler knows its revenue.
The free leak snapshot reviews a limited set of public pages and reports observable link behaviour. Fit for deeper work is reviewed separately.
Get a free leak snapshot