For adults 18+. HelloViolet publishes reviews of third-party AI services and may earn a commission through links on its site. We used our own review for this experiment because we can identify both the link a reader sees and the financial relationship behind it.

Two buttons, one destination

What does a reader know at the moment a review offers to take them to a product? On November 10, 2026, we saved the HTML and stylesheet served by our Candy AI review. In the quick verdict near the top, a button invites the reader to try the service. Another button near the end makes the same offer. The two links have the same label and the same complete destination address. They give the reader two opportunities to leave the page, not two different offers.

Both links carry the HTML relationship value sponsored. The buttons themselves do not display that word. The page does explain, in ordinary language, that the publisher may earn a commission if a reader subscribes. But that paragraph comes after both buttons. Reading order matters here: a person can reasonably follow the review's first invitation without ever reaching its explanation.

What was on the first screen?

We rendered the saved page in Chromium at two viewport sizes. We used the page's own stylesheet, started at the top, left the browser at its default zoom and blocked external requests while rendering. We measured from the top of the page to the top edge of each element, in CSS pixels. We didn't open either offer, create an account or inspect a checkout.

Browser viewportFirst offerSecond offerCommission notice
390 × 8445462,2062,592
1,280 × 8004441,6842,070

On the narrow layout the first button starts inside the initial 844-pixel screen. The notice starts 2,046 pixels below the button. The wide layout fits more text on a line, but the order stays the same. The first button starts inside the initial 800-pixel screen; the disclosure is still below the second button. These are measurements of two renders of one dated page. A larger default font, another browser or a later edit could give a reader different numbers.

Change one thing and measure it again

In a local copy we placed a second copy of the existing disclosure paragraph immediately before the first button. We kept the original notice near the end. The wording, button label, link address, fonts and viewport dimensions were unchanged. That isolates a fairly small editorial choice: when does the reader get the explanation?

On the narrow layout the new notice starts at 552 pixels and the first button starts at 637. On the wide layout the corresponding positions are 450 and 513. In both renders the notice now precedes the first invitation in the document as well as on screen. The first button still begins within the initial viewport. For this particular page, giving context earlier did not require removing the quick verdict or moving its invitation to another screen. The rearrangement was an offline experiment; those latter positions do not describe the live review.

A check you can run on another review

Start from the place where a reader would make a decision. Pick the first link that leaves the review for the product, then read only from the headline down to that link. Can you tell who wrote the recommendation, who operates the product and whether the publisher can benefit if you buy? Next, read the rest of the page. If you find an explanation later, note which earlier decision it would have helped. Counting disclosures on the whole page would miss this distinction.

For an editor comparing two layouts, keep the browser width and the article text fixed. Record the location of the decision and its explanation in both versions, including the order in the underlying text. A shorter gap by itself can mislead: moving the notice to ten pixels below a button still puts the invitation first. For a reader, the useful question is simpler: what did the page tell you before you had to choose?