Will SearchGPT Whack Google Where It Counts?

For the last 25 years, people have used internet search in a straightforward way.

Input a keyword or phrase, and the search engine returns pages and pages of results that match identically or conceptually to the request. The results were ranked by relevance — the word that became the Holy Grail for getting the coveted No. 1 slot on the results page.

Marketers cried everywhere, “How do we become more relevant?” and “How do we rank?”

The answers would deliver the prized organic — also known as free — traffic to your brand’s content.

Well, a week or so ago, the next big step in search arrived. It’s not really an evolution; it’s more like a devolution of how people can find information on the internet.

OpenAI launched ChatGPT as an AI-powered web search engine, bringing real-time information directly into conversations. As of this week, paid subscribers and people on the SearchGPT waitlist can access the new capabilities. Free, enterprise, and educational users will get the new feature in the coming weeks.

What will it do to traditional search? Should this change your cries about relevancy and ranking? Will you now ask, “How do we become THE answer?” and “How do we get cited?

Instead of turning to the chatbot to find out, we asked Robert Rose, CMI’s chief strategy advisor, for his take. Read on or watch this:

Reflection on search early days

The early days of internet search centered on the giant that is Google. It set the standard for everything.

Ever wondered why Google’s search results page was so ugly? Why did they never use images to break up the page or publish icons to help navigation? Google wanted to make searches as fast as possible. For years, Google highlighted how reducing its logo’s file size on search results pages saved bandwidth and, therefore, energy requirements. That’s how big Google was.

Google’s power was so great that Googlewhack was a game in the early days. Google’s index was so big that most combinations of words would return many results. So, the game was to come up with a two-word combination that would return exactly one result.

My favorite winner? “Demurrable insufficiencies.”

Googlewhack ended about 15 years ago when search indexing changed.

My point is that, like the size of the universe, it’s hard to understand how comprehensive, popular, and ubiquitous Google search has become and its influence on how people access information.

But AI search is growing. Google has begun including AI-generated overviews as one component of its classic search engine page. And, over a week ago, OpenAI launched an integrated web search into ChatGPT’s interface.

The ChatGPT feature determines when users can click or tap into web results based on queries, though they can also manually trigger web searches. This addition closes a key competitive gap with rivals like Copilot and Gemini that have long offered real-time access, which isn’t surprising given that their companies — Microsoft and Google — are behind two of the most popular search engines.

Is this the future vision of AI search?

Is OpenAI defining AI search the same way Google did for web search? Will it be better? Is it better already?

As a user, I must say it’s darn impressive. I used the ChatGPT Chrome extension as my default search browser and made my common everyday queries such as “marketing news this week” and “10-day forecast for my area.” It returned good results.

Further, the “answer” appears in the window, and the query conveniently goes on the left-side activity tracker. The sources cited in the answer appear on the right, followed by general search results.

Now, this is the interesting part. As far as I can tell, it cites three to five sources and provides another 12 results. That’s it — no page after page of results. Now, some may find that bad, and some may find it good. But one thing is certain: If ChatGPT keeps that format and its search takes off, a battle royale will ensue for those 12 spots listed for a query.

Here’s the other kicker. Articles have already been published about how OpenAI finds, indexes, and ranks content for display. Spoiler alert: It’s the Bing index, which makes sense given OpenAI’s partnership with Microsoft. But OpenAI also uses its own indexing bot.

I discovered the results for my search tests on Bing did NOT have the same rankings as those on ChatGPT. For example, a search for “workflow tools” on ChatGPT turned up a Techopedia listing as the fourth organic result. On Bing, that result appears on page two — about the 15th listing.

When I made a subtle change to the ChatGPT query, “OK, now I’m serious. Can you tell me the real best workflow tools?” ChatGPT returned very similar results, and only the order differed slightly.

However, when I changed the ChatGPT query to “I need a workflow tool for marketing content — which one is best?” ChatGPT gave me different answers altogether.

Welcome to the next round of search engine optimization. How do you become THE answer? How do you become ONE of the cited answers, AND how do you become a generic search result? Of course, I suspect Open AI will quickly monetize this so you can purchase your way into many of these recommendations and lists.

Will AI search be an easy game?

Finally, I decided to see how ChatGPT would handle a game of Googlewhack, or, maybe I should say, Chatwhack. I tried my old winning combo of “demurrable insufficiencies” — just two words meaninglessly stuck together.

Today, Google returned exactly 15 results, which all reference the Googlewhack game. ChatGPT desperately wanted to create an answer.

Instead of saying, “You’re joking” or “Come on, that’s not a real word,” ChatGPT said, “demurrable insufficiencies likely refers to deficiencies in a legal pleading that render it susceptible to a demurrer. In California, for instance, a demurrer can challenge a complaint on grounds like lack of jurisdiction, lack of legal capacity to sue, or failure to state facts sufficient to constitute a cause of action …” It also cited a few sources that define demurrable and insufficiency.

Uh, OK.

This Chatwhack test illustrates an interesting thing you must watch out for: The biggest insight may be what you don’t get from search results. What does it say when you ask for insight on a particular restaurant or pose a technical question and get 15 nonsensical results from the internet?

Even if it makes up desperate responses, ChatGPT must achieve its goal of providing an answer. In some cases, these forced answers are likely to be combinations of sources referring to one part of the query or another.  

So, remember, searchers will get an answer, but it might not be the answer.

Big adventures on the way for search, marketing, and AI — and ChatGPT might just be the tipping point that brings you to it.