In the next several years, it’s quite possible that the way we think about keyword targeting within the SERP may undergo a significant, perhaps radical transformation; even to the point where a good portion of the time the ads being triggered against keyword queries bear little resemblance to the keyword phrases in our accounts, yet are still highly relevant to the searcher. Google’s adoption of “semantic search” may be one of the primary drivers of this change.
The search marketing community was all abuzz recently when Google announced via a Wall Street Journal article that stated:
“Google Inc. GOOG is giving its tried-and-true Web-search formula a makeover as it tries to fix the shortcomings of today’s technology and maintain its dominant market share…
Over the next few months, Google’s search engine will begin spitting out more than a list of blue Web links. It will also present more facts and direct answers to queries at the top of the search-results page.”
The phrase “semantic search” gets thrown around a lot. But in reality, the evolution toward search engines including Google imputing intent of the searcher in a semantic or behavioral way has been around for a long time. For example:
- Google imputes location-based intent without the user specifically requesting local results.
- Google guesses when to trigger universal search results including most recently real-time results, but also in the distant past other universals results including:
- Shipping info links (FedEx and UPS)
- Google runs AdWords ads based on “session-based ad targeting” where the keyword in your campaign was used during the same session but not on the specific SERP.
This last example of session-based ads may actually be the closest approximation to how the SERP and the ads triggered for a specific search will evolve when semantic technology is applied. From my research, many of the user search instances that trigger a session-based impression for a prior keyword happen when the subsequent keyword search resembles the intent of the searcher while using a completely new set of keywords.
Perhaps a new match type will be invented or perhaps semantic ad targeting will be covered under “expanded broad match,” which can get fairly broad already. However, I predict that as Google refines its semantic prediction engine, we’ll see a whole lot more impressions under this or some new match type that behaves more like session-based advertising matches, where a look at the keyword will leave many of us scratching our heads. Unfortunately, if session-based broad match is to be any guide, we won’t be able to opt out of semantic matching, and similarly, I expect conversion rates to be 15 to 35 percent lower than a traditional broad match and I also don’t expect Google to enable internal “smart pricing” (yes, I did cover smart pricing in 2004), although Google should do so on both semantic- and session-based search advertising.
However, the news isn’t all bad. Many advertisers struggle with limited inventory in keyword lists, and because semantic search should allow Google and other engines to better understand the true user intent of a search query, the inventory available to get clicks rises with perhaps only a small ROI hit.
Google may decide to run landing page content though a semantic engine to assist in picking the right ad for each search, ads you don’t expect to see run. In the short term, Google is making sure users still feel that Google has the lead in providing the best results, because if users decide that Bing or Siri, for that matter, provide better results, Google loses its most valuable asset, the searcher.
In the meantime, it probably puts even more pressure on advertisers and marketers to make sure that their landing pages (for ads as well as site pages in general) are well written and descriptive. Surrounding content can help disambiguate the essence and key concepts of a page. For example, a page with the word “apple” on it and the word “cook” or “bite” would more likely be about a fruit, while a page with the word “apple” on it that also includes the words “display” or “memory” are likely to be about a computer.
Amit Singhal at Google mentions “Freebase, an open-source knowledge graph” in his follow-up to all the commotion surrounding the WSJ article. Clearly, that technology will help enable Google to impute intent better than it does now.
If Google and Bing get good at semantic search, we may actually see a shift back to natural language search queries. Most users learn not to use those queries now because the results tend to suck. If that’s the case, then we may want to adapt both our content and keyword strategies.