Actually, a different study than the one linked by the OP states this. A study they are careful not to elaborate upon in this press release. I worked for StarCom 1997-2000 in their new & shiny online advertising group. StarCom is heavily incented to produce research proving the effectiveness of all types of advertising. And they own a much larger share of online display advertising than they do of paid search engine marketing. I can't prove that their current research is biased. But when I worked there, we were quite careful to present our results in the best possible light. For example, we ran a very large banner ad campaign for a Proctor & Gamble product. There were two reasonable ways to determine if it was effective: Ask an online panel if the recalled seeing the banner ads in question. Or look at the actual grocery store / drug store sales data to see if the banner campaign caused any lift in sales. The first argued that the millions spent on this campaign produced OK brand awareness; the second demonstrated that it did nothing to move the product. Guess which of these two results went into our sales and marketing materials and which was omitted.