SearchGPT vs. Other Search Engines: A Critical Comparison
At the beginning of August, OpenAI (makers of ChatGPT) launched its beta version of SearchGPT to compete directly with the likes of Google Search and PerplexityAI. With its inception, we were eager to learn about the new technology, explore its differences, and see how search as we know it may be altered in the immediate future.
We put OpenAI’s SearchGPT in the hands of our SEO experts. Let’s delve into the findings, key differences, and limitations, as well as the long-lasting effects AI engines may have on search.
First Impressions of SearchGPT
- It’s a mixed bag. It performs well on some queries, and poorly on others.
- There is no apparent rhyme or reason to ranking.
- If an algorithm is being applied, we can’t tell what the rules of engagement are.
- Results lack diversity.
- Results are completely different if you re-search a query. Content is different, format is often different.
- You might get sources, you might not. 🤷
Performs Well on Technical & Informational Queries
When inputting search prompts requesting technical information, like “How are petrochemicals used for petrochemical producers,” SearchGPT performed well. The platform provided a brief summary of how different petrochemicals are used in general.
“how are petrochemicals used for petrochemical producers” – take 1, SearchGPT
The result was similar to what Google’s AIO provided:
“how are petrochemicals used for petrochemical producers” – Google
The big differences are:
- SearchGPT provides thumbnails, Google does not.
- Google provides sources. SearchGPT does not.
In fact, there are sources for the thumbnails provided by SearchGPT, but not for the text information it is displaying. When we re-performed this search, we were given sources… weird:
“how are petrochemicals used for petrochemical producers” – take 2, SearchGPT
For many of our detailed prompts, looking for technical information or asking related questions, SearchGPT provided valuable insights and comprehensive material. The results were fairly robust, appeared to have been pulled from a number of sources (if they appeared), and were similar to answers given by Google Generative AI.
Performs Poorly on Commercial Queries
Additional search prompts gave us a lackluster result. When we searched for “dog sweaters,” intentionally a commercial query, we expected to get a link to click, like we typically would in a Google search. Instead, we were greeted with images of dogs in sweaters (an informational result).
“dog sweaters” – SearchGPT
While cute, the images do not satisfy the commercial intent of the query. Further, the text results seemed to reference specific products, and their availability at specific merchants, but without any way to click-through and purchase. Turns out, the list of results does correspond with the images above the list – but navigating this is a lot to ask of a user.
Does this indicate a need for image optimization to be more at the forefront for product based businesses? Or a pretty big bug to work out before users will find it useful?
The other key takeaway is, since the platform is so new, it may be good practice in the immediate future to verify source data and ensure that information is accurate and not misleading. That is, if SearchGPT graces you with its sources.
SearchGPT & Local Search
When searching for localized businesses, SearchGPT showed significant limitations when compared to the likes of Google or Bing. In traditional search, businesses are ordered based on factors like reviews, proximity, and whether they have a website. In our experiment, SearchGPT doesn’t seem to follow a structured approach when searching for a local business. There are no sponsored ads and no reviews associated with businesses. Many of the businesses didn’t have a website, and the businesses with a website often had thin content on its pages.
It is worth noting that SearchGPT cites Google Maps as its only source for results, but displays them in an entirely different order from a search performed in Google or Google Maps.
“AC repair near me” – SearchGPT
You’ll also notice that SearchGPT treats “near me” and localized searches differently:
“AC repair Orlando” – SearchGPT
In the localized results, the websites SearchGPT served had thin content, services information confined to one singular page (lack of depth), and an overall lackluster user experience — something that would hurt rankings on a traditional search engine. We even came across a homepage that returned a 404 error.
A key difference between traditional search and SearchGPT is that popular local websites like Yelp are noticeably absent, suggesting that the tool might not yet incorporate the extensive databases and user-generated content included in traditional SERPs.
This lack of robust filtering and apparent ranking factors indicate that SearchGPT’s “algorithm” operates under a completely different set of rules, almost as if it disregards the best practices that have shaped traditional search optimization.
How Does SearchGPT Compare to Other Search Engines?
When testing SearchGPT, our team noticed a healthy set of similarities, and also some stark differences versus other search engines.
SearchGPT vs. Google
So, how do SearchGPT’s search results compare to Google? As you can see, they really don’t. For products and local services, SearchGPT doesn’t seem to have a tried and true algorithm yet. In some ways, it has the feel of a paper catalog or phonebook. The links in the product results don’t take you to the website; they instead take you to an image.
Local search in SearchGPT does assemble local businesses, but it appears that the ranking of the results needs to inmprove the user experience. Google, and to a lesser extent Bing, have mastered the art of providing the best user experience for their searchers. This has been done over time through changing algorithms and countless SEO and search experts optimizing websites that, in turn, improve search results.
In short, SearchGPT offers a new look and feel when compared to traditional search, but a poor user experience. It doesn’t seem to utilize the same “best practices model” that traditional search has demanded and rewarded. There is a lot that needs to be addressed in the near term if SearchGPT is to compete.
SearchGPT vs. Google Gemini
In 2023, Google, in an effort to compete directly with other AI platforms, created the generative AI feature at the top of the SERP, dubbed the AI Overview (AIO). Some like this feature, and others dislike it. One thing is for certain: its results pale in comparison to other AI models, often providing less detailed information but concisely answering the question. It’s certainly aimed at being a helpful feature, but it’s new for Google; they’ve built their reputation on organic search results.
In comparison to Google AIOs, the results were much more detailed and provided multiple sources, and we walked away feeling more confident with the information being served by SearchGPT. While it also provides a list of resources that offer additional information from reputable sources, Google’s AI Overview provides much shorter summaries that give only high-level details. For AIO purposes, SearchGPT delivers more.
SearchGPT vs. Perplexity
The platform that was most similar to SearchGPT was PerplexityAI. When prompted with the same query, Perplexity gave a similar output to SearchGPT, providing detailed information and an abundance of resources.
“how are petrochemicals used for petrochemical producers” – Perplexity
Perplexity, at the moment, offers the best of both worlds, where it provides the comprehensive details to answer our prompts in the engine but also provides opportunities to view numerous additional resources to really home in on answering the prompt, which is similar to the experience of what traditional search has provided us in recent years. Plus, it looks pretty sharp.
SearchGPT’s Growing Pains
It seems that SearchGPT operates in a different manner than traditional search engines. Currently, it prioritizes less structured, experience-based content over the refined and authoritative information that Google provides. SearchGPT conclusively lacks the sophistication in its algorithms that Google is known for, resulting in what we consider a less reliable search experience.
Ranking First, and 2nd, and 3rd…
One of the notable issues with SearchGPT is the duplication of resources. For example, we searched for “red Nike running shoes,” the tool provided four options, all from the same retailer — Dick’s Sporting Goods. This kind of repetition reduces the variety and depth of information available to users.
Inconsistent Results
If you input a query into SearchGPT more than once, you get different answers, nearly every time. This is further evidence that there is no consistent system, process, or logic applied to ranking.
“dining room chandeliers” – take 1, 2, and 3, SearchGPT
Is Metadata a Ranking Factor at all?
Additionally, SearchGPT handles links and metadata differently from Google. When you click the “link” in SearchGPT, it lists the websites used for its search results, but these don’t appear in the same format as Google’s title and description tags. Instead of using metadata comprehensively, SearchGPT relies heavily on the H1 tags, which limits the depth and precision of the information presented, which users use to decide where to click.
Can We Optimize for SearchGPT?
The short answer to whether or not we can optimize for SearchGPT is yes.
What Will be Important for SearchGPT SEO?
One important factor to remember is that SearchGPT is a service provider, and its ultimate goal is to provide value in its service. OpenAI’s services tend to offer free and subscription-based entry points, which indicates that it could choose to monetize SearchGPT by offering different search features to different users. Google’s business model is completely different, being monetized by ad revenue.
While it remains to be seen how SearchGPT will monetize, it most certainly will in some way. Things like UX/UI will need to be at the forefront of SearchGPT’s algorithm as it evolves.
We believe that tried and true SEO methods related to user experience and content relevance will likely still apply to SearchGPT SEO. Similar to the SEO shift with Google in 2022, the focus of optimization should be on helpful content. The content should be focused on answering a user’s questions and queries concisely.
Another opportunity to optimize for AI search is by incorporating testimonials or real-life experiences into content. Any opportunity to provide insights into products, services, or opinion-related content will be an opportunity to show up in AI search results.
What Might the Future of Search Engines Look Like?
Search is poised to rapidly evolve. AI engines like Perplexity and SearchGPT already provide an entirely different experience than traditional search engines.
If SearchGPT and other AI models seek to gain significant market share, they will likely need to continue creating a search experience that is far different than Google and Bing’s traditional UX. In order to do that, they will need to focus on the quality of the results and user experience. Google has invested decades and countless dollars in the UX of search, and its users span multiple generations. SearchGPT will need to provide something more than the novelty of AI to capture and keep users.
SearchGPT and SEO: Time Will Tell
SearchGPT has good potential as a search engine in time, but it has kinks to work out. Google dominates the market share of search, and in order for SearchGPT to gain a slice of that pie, it will need to provide quality results across all types of queries.
As businesses consider whether to begin optimizing for SearchGPT, it’s crucial to consider the pros, cons, and key differences. While proven SEO practices remain the here and now, there are always new strategies and A/B testing to explore. For SearchGPT optimization, there may be opportunities particularly regarding visual content and ensuring that on-page content is structured to best align with how the platform operates (although we wouldn’t recommend this today).
In our professional opinion, we don’t see SearchGPT upsetting the apple cart soon. While there could be nuances to rank in SearchGPT in the future, the core tenets of SEO hold true: make exceptional content to answer your user’s intent. And optimize for AIOs as part of your user’s search journey.
It’s clear that SearchGPT is still in its infancy, and its best practice guidelines and algorithms are certain to evolve as the technology ages.