AI Driven Growth

AI forms a core part of our strategy in the organic growth of your search engine rankings. Read on to find out how we do this.

There is no question as to whether AI has revolutionized our world. In just a few short years, we have seen many jobs that were considered outside the realm of automation become automated.

Traditional automation methods, which rely on custom logic and conditions, have been superseded by algorithms that are able to be presented with problems, simply told what the solutions are, and are then able to work out everything in-between by themselves.

These neural networks are able to make accurate future predictions on data they haven’t seen before. This is of course an oversimplified explanation, but it is at the heart of what AI has become.

Q-Learning In Games

Modern AI (neural network based AI) has been around for many decades, however it hasn’t been until recently that the field has boomed with some revolutionary new research. In 2015, a landmark research paper was released by Mnih. et al. “Human-level control through deep reinforcement learning”.

This paper built upon earlier research which demonstrated the ability of reinforcement learning agents, implementing an algorithm known as “Q-Learning”, to play Atari video games better than human players.

Whilst AI playing video games doesn’t seem that interesting to most people, it is a big deal in the field of AI since most  problems in the world can be modelled as an agent attempting to make the right decisions to solve a problem. Video games are perfectly suited to simulate this.


After the landmark paper by Mnih et al., Google confirmed the implementation of their new RankBrain algorithm which utilizes AI to deliver optimal search results. The two events were most likely not linked at all, however the timing is interesting as it was at this point in 2015 that we can observe AI began its modern trajectory of becoming incredibly efficient at performing tasks that it was widely considered only humans could do beforehand.

How is this relevant to you, us, and your rankings? It is because RankBrain, and the new search engine ranking ethos, changed completely in 2015.

Gone are the days of simply tricking the latest search algorithm with a new link building strategy that circumvents it. The thing with modern AI is that it can be incredibly efficient at certain things – even better than us humans.

Identifying spam is one thing that it’s incredibly good at as can be seen with the simple yet effective Google Captcha algorithm.

Search AI & GANs

If modern AI is able to identify spam well, then it isn’t a big leap to imagine that it’s able to also identify search engine spam well, including things such as spam links pointing to a website. Whilst it’s true that even today it’s still possible to temporarily fool search algorithms, doing so is an ultimately fruitless exercise since the algorithms are constantly self-learning and expanding in real time, as opposed to using fixed updates like before.

This is because a lot of any search engine algorithm is now automated and self learning as opposed to requiring a human to manually update it like before. As the AI identifies more real and fake link building techniques, it becomes even more efficient at identifying new types of spam.

Lately, we’ve observed that an interesting new field in AI seems very relevant in identifying both efficient and spammy backlink building techniques. Enter into the realm of Generative Adversarial Networks. Click on that link for a brief introduction to what GANs are, straight from the horses mouth (Google).

TensorFlow 2 is an AI framework written by Google that we use internally for all our search engine simulations. Perhaps it’s a bit ironic that we use software written by Google itself to get a better idea at what’s more effective at ranking in their search engine, but it just so happens that Google is also a technology leader in AI. So is Microsoft, the owner of the 2nd largest search engine (Bing).

If Google and Microsoft are not only aware of the latest research in AI but are indeed pioneers of AI research themselves, it’s a no-brainer that they’re also using the latest AI methods to optimize their most profitable product: search.

GANs are effective because they are based upon training a neural network to identify real or fake things. This can be anything from identifying whether an image is a photograph or hand drawn. As the network is exposed to more and more examples of both real and fake things, and gets more and more feedback as to how it’s doing in identifying the difference between them, it becomes better at doing its job.

Eventually it can distinguish the difference between the two things even better than humans can. There are of course exceptions; for instance, whilst self driving cars are able to react far faster than any human can, they are also fooled by simple things that we would consider trivial.

Indeed, AI is better than humans at the board games Chess & Go, however if you worked out how to counter the specific algorithm that it came up with, you could fool it with things that we as humans would consider easy and trivial, even though the AI is much better than us at those games. It’s important to understand that when people say AI is “doing some things better than humans” we really mean specific things within that field, not everything.

One thing that AI can now do better than humans is identifying search engine spam at scale. An AI is able to pool information from potentially millions of different signals. Cruder, earlier search algorithms relied upon “signal indicators” that search engine engineers would manually program.

These algorithms usually consisted of a few hundred signals. We believe search algorithms are no longer limited to hundreds of signals anymore, but potentially millions, and that attempting to identify all these search signals and playing around each of them manually is no longer an effective SEO strategy, at least not since 2017.

So what is an effective SEO strategy today? Since 2017, we’ve gone all-in on AI, particularly GANs and reinforcement learning agents (the AI used to play video games). We write our own internal agents and train them on identifying the difference between a good backlink and a bad backlink, based upon many signals that not even a top statistician could effectively find correlations between, but that an AI can.

We then test these AIs in live environments to see how well they do, and whether they can identify types of backlinks they haven’t seen before. So far, they’ve proven effective, and in fact we believe search engines are probably doing something very similar to what we do.

For all our business and enterprise clients, we put together a custom growth plan that is based upon signals that our own internal AIs give us. These signals are based upon unique factors within your industry or niche, such as competing sites and the types of backlinks that cause them to rank well. We simulate our intended SEO strategy through our proprietary GAN algorithms to identify whether any of the link building techniques we or our AIs come up with could be considered spammy or not. If a competitor makes a ranking mistake, this is something we can capitalize on by presenting your website in a more favorable manner to the search engines.

We do this as part of all our Tier 2 – Tier 4 link building strategies. For our Tier 1 link building, we use the AI less, instead preferring to rely on more qualitative methods such as manual outreach, connections, and reputation building. Nevertheless, any effective SEO strategy requires a combination of both qualitative and quantitative methods.

The combination of our years of intuition in search optimization with our own internal algorithms that are based upon the latest research in AI, allows us to provide both these things in a comprehensive, custom growth plan that aims to meet your search ranking goals.

We recommend you give us a try for 6 months and see for yourself the results.