Meet BERT - Google's New Language Detection Algorithm

Social media marketers, it's time to meet Google BERT (or "Bidirectional Encoder Representations from Transformers" in technical terms).

BERT isn't just your normal marketing contact. In fact, he's not a human contact at all. Instead, the Google BERT update represents one of the most important changes to the way Google ranks content in the past decade. And everyone who relies on search rankings will need to know how it works.

What is Google BERT?

Basically, BERT is Google's latest attempt to match up the meaning of search queries with the most relevant results. It uses a technique called Natural Language Processing which itself employs AI to understand what searchers are looking for, and how well content matches up to their needs.

This can result in subtle changes to the way searches work. Small modifiers in queries such as "not", "to", or "no" may previously have been discounted by Google's algorithms, which took a broad-brush approach. Not any longer. When BERT is fully rolled out, queries will be assessed in much more detail, teasing out nuances that may have been missed before.

It's important to note that BERT isn't a replacement for RankBrain - which was introduced in 2015. It's probably better to see BERT as an add-on which will kick in for certain queries where its capabilities are most useful. At the moment, experts reckon this applies to 10% of Google searchqueries - which is still a massive amount of search activity.

How does Google NLP work?

NLP lies at the heart of BERT, so it's useful to know a little about what it does, and why this matters from a digital marketing perspective. The term "natural language" is a bit misleading here, as NLP is actually an attempt to create AI simulations of the way people actually write and speak.

Using neural networks, it tries to analyse the structure of text, which isn't always the same as applying strict grammatical rules. A process of pre-training exposes BERT to billions of questions, which it dissects and categorises, putting every word into multiple contexts. It also employs a "bidirectional" way of doing so, allowing Google to get to grips with complex queries, across almost all cultures.

What the Google BERT update means for online marketers

This is the million (or billion) dollar question. Whenever Google releases an algorithm update, it causes a certain amount of stress for marketers, who are not sure how well their content will score. And we can't tell for certain how BERT will play out, but some things seem likely.

1. Testing and updating content will be essential

Firstly, as with all major Google updates, it's advisable to audit existing content to see whether there have been any major downgrades (or upgrades). Check weekly traffic changes for different sites, and try to extract some insights about which pages are suffering, and which ones are thriving.

When you have a better idea of the state of play, you can formulate content strategies to fit BERT's requirements. In that case, reaching out to native copywriters with the ability to create natural-sounding, relevant content seems like a good route to take.

2. Clarity and relevance are more important than ever

On a related note, the way NLP works suggests that Google will increasingly privilege content that is on point and as relevant as possible to the needs of searchers.

This means that flabby content which talks around the point instead of directly addressing the requirements of visitors, won't rank well. There's been a trend towards focusing on relevance and the delivery of actionable information for some time. However, BERT reinforces this, and may well leave poorly written, keyword-stuffed pages languishing at the bottom of the content pit.

3. BERT doesn't mean the human touch is irrelevant

With the rise of AI-assisted searching, some commentators have argued that human writers will soon be replaced by computer-generated alternatives. But the truth is that AI is being used to filter out high-quality human writing from irrelevant or nonsensical content.

Moreover, human writing will fit into the concept behind BERT, which interrogates the way people actually write and speak. BERT is designed to work around humanity, rather than providing a machine-based upgrade, and it seems unlikely that robotic, artificial texts will score well.

Even so, while machines will not replace copywriters any time soon, it's essential to focus on efficient writing and practical information. Think of it as a sharp reminder that content creators need to stay on their game.

Conclusion: Do we need to fear BERT?

With luck, content marketers will find a way to thrive with BERT's assistance. Google's new algorithm tweak will alarm some, but inspire others to create better content that serves customers more effectively, and that can only be a good thing.

If you need some assistance in crafting content that matches the needs of NLP, and that BERT adores, greatcontent can help. Get in touch with our team, and we'll help you publish texts that harness the power of Google's engine, while keeping pace with its changing requirements.

Text: Sam Urquhart
Image: searchengineland.com