Insights
Rethinking Search
What eventually contributes to Google's decline won't be another search engine.
It will be something that encourages you to use Google search less–or perhaps any 'search engine.'
Last week, I came across a startup that is rethinking search from the ground up.
They're called Exa and they just launched Websets.
Their goal is to transform the messy web we know into a structured database.
Here's my understanding:
- Websets are not based on keyword searches. Exa converts webpage content into embeddings that capture meaning rather than just matching words or phrases.
- The company has processed around 1B web pages so far. Google has about 1T pages, but Exa is prioritizing depth of understanding.
- The most exciting thing about Websets is that it makes extremely complex searches viable.
For example, you could ask for highly-specific lists of people or businesses:
"Find me technical SEOs based in the Bay Area who've worked at startups and also in AI."
After entering a search you might have time to make a coffee, but the results should be worth it. You'll get precise lists containing hundreds or thousands of entries.
If you need proof, check out this oddly specific sample from Exa:
LinkedIn Profiles: Male; finance; Northern European last name; athletic.
It may be slow, but this database-style approach could reshape how we find and organize information on the web.
Imagine how you could use it.
The possibilities are endless.
Exa say they want to "build perfect search."
At Exa, our goal is to build perfect search — i.e. turn the internet into a super powerful database that you can query however you want.
This is a big first step.
That first step is an encouraging beginning.
If you want more info, you can watch the founder as he demos Websets and/or signup to their waitlist.
SEO in the Age of SearchGPT
OpenAI is gradually building a self-contained location to keep users on its platform.
That’s a logical move given that ~40% of their users exit to Google (according to Semrush data):
Since the integration of SearchGPT (for some users), ChatGPT is positioning itself as a search engine.
This move seems to have been well planned.
OpenAI has formed numerous partnerships with media companies, and recent reports suggest they are working on a web browser with ex-Google Chrome developers.
While SearchGPT is not a direct competitor to Google now, it presents new considerations for SEOs.
Let’s quickly review:
Content creation
SearchGPT is designed for conversational interactions, focusing on questions, context, and discussions.
This means we must create content that reads smoothly, answers questions directly, and provides concise information.
For example, incorporating FAQs and focusing on clearly defined terms and definitions will cater to SearchGPT’s conversational nature.
Use a variety of media
SearchGPT can reference and describe different types of media, not just text.
If you add videos, infographics, and interactive elements into your content this will improve its value.
Label your multimedia content
SearchGPT does understand images, but descriptive text is still important.
It makes your content inclusive and accessible to all end users. Don’t ignore this.
Target high-visibility content sources
SearchGPT seems to link out to content creators more than Google. You could think about targeting high-visibility sources used by SearchGPT to increase your brand’s visibility.
Remember: Always meet user intent.
Keep your content up-to-date
SearchGPT (like Google) prioritizes relevant, up-to-date information. Regularly review your content to keep it fresh.
In the end, the future of SEO will likely involve a hybrid approach, balancing optimization for both traditional search engines and emerging AI platforms like SearchGPT.
A proviso
I’m not suggesting you rely on SearchGPT for research.
It continues to make things up, provide wrong answers, cite incorrect sources, and offer inconsistent answers.
Always check your sources before publishing content, and verify that you’re linking to an original article.
If you want facts, go to Wikipedia!
Huge Google Search Study Reveals How Users Behave
A new study by SparkToro and Datos found that just 148 keywords make up 15% of all Google searches.
The data also showed that:
- 10,000 queries make up nearly 46% of all search activity.
- 44% of searches are for branded terms.
- 51% of searches are informational, like weather updates or “how-to” questions.
- 14.5% are commercial, aiming to find products or services.
- Only 0.69% of searches are transactional, i.e. when someone is ready to buy.
Does that spell bad news for direct sales?
You can dive into all the details or just check out the key takeaways from Rand’s post.
Additional resource
- Listen to what Sam Altman says about Google, ChatGPT, and search in this interview. Start at 1.18.00.
- Go Fish Digital observed that for the query "digital PR agency," ChatGPT Search was not listing "Notable Clients" for their company — unlike their competitors.
They also realized that ChatGPT Search was frequently citing an article on their website. Chris Long explains what they did and how they fixed the problem.
I hope these insights help. For more tips, follow me here.
