After 20 years, search engines are starting to look different

After 20 years, search engines are starting to look different

Reviewr Team

Reviewr Team

Search Evolution: Today and Tomorrow

Search engines have more or less functioned the same over the past 20 years: recommending web pages most related to search queries. New AI-powered language models, like GPT-3, are on the trajectory to radically change this experience.

The current state of search is doing a better job of parsing and organizing web content.

neeva, best backpacks search

For example, see how (an emerging ad-free Google competitor) does this for shopping searches. This “best backpacks” search shows the products expert sources recommend, without the need to visit any pages.

google backpacks, best backpacks search

Google has rolled out a similar feature that consolidates content into a filterable product slider with supporting specs and reviews.

These experiences are currently limited to specific regions and the most searched product categories; however, this trend expects to continue more broadly.

The next leaps in search expect to provide more thoughtful information natively. Results may start to look similar to Wikipedia, where the outcome is a unified response rather than a ranked list of web pages.

google report
How expert-grade answers from Google could look like

Researchers from Google recently released a paper that proposes Google use advanced language models to provide expert-grade answers, rather than just being a gateway to possible answers.

google model
Current vs. model-based Google search schematics

The paper proposes a change from “retrieve-then-rank” to “unified retrieve-and-rank”.

This study outlines five critical factors required for this method to be successful: authoritative, transparent, unbiased, diverse perspectives, and accessible. To get this all right, rolling it out on a grand scale will take some time.

In the meantime, GPT-3-powered companies like are achieving similar results at a smaller scale.

baqpa by uses GPT-3 and other language models to collect and summarize review content. In this case, aggregated reviews of the most popular backpacks. Reviews are scored based on their “helpfulness” and “positivity” and then summarized using GPT-3 to provide a democratic consensus of each product.

What does this mean for discovery?

1. Content depth increases in importance

As search engines become better at providing detailed answers to complex questions, content is still needed to contribute to those answers. Without source content, search engines cannot educate themselves to give answers.

However, content with surface-level depth will see a significant drop in traffic. If content provides an in-depth explanation of a question and a detailed answer is required, traffic levels should be far less impacted.

The same goes for various sophisticated tools and other curated resources that search engines will not host natively.

2. Brand reputation increases in importance

Currently, reviews of businesses are scattered across the internet and only ever summed up by star reviews.

A more articulate unified voice will emerge, so the language people use to describe things will become more impactful than ever.

This unified voice will allow people to find out the good, bad, and ugly about anything faster than ever.

Sweeping things under the rug will become less of an option. The best that can be done here is to improve customer satisfaction continuously.

3. ???

While this is a gradual transformation, a lot of changes are anticipated with these new advances. Tweet us at @reviewr_ai and let us know your thoughts.