How AI is Changing Paid Search

Artificial Intelligence opened the door to a new level of automation and optimisation in paid search. Organisations can now manage large-scale advertising campaigns far more efficiently than they ever could have done manually through functions such as smart bidding.

In fact, a case study by Google of India’s travel booking engine Goibio showed that Smart Bidding increased hotel transactions by 25% at a 22% lower cost per conversion for hotel non-brand search campaigns. In addition, using one-click automated upselling online ads has shown to increase average order size by more than 360%.

Despite the benefits that these types of automation bring, there is a possibility that paid search is going to be transformed forever and that advertising itself may become less relevant.

This article reviews some of the ways in which AI is currently implemented in paid search and some possible developments that may render the paid search landscape unrecognisable in future.

Smart Bidding

Google and other providers allow users to select Smart Bidding for their campaigns so that bids are calculated automatically. Google has five bidding strategies to choose from depending on the business’ goals: maximum clicks, maximum conversions, target CPA, target ROAS, and target impression share. (This highlights how humans are needed in order to define relevant KPIs based on specific goals.)

The introduction of smart bidding made it possible for organisations to manage hundreds of campaigns with ease, at a lower cost and higher CTR. Emerging technologies may influence the role of smart bidding, which this article will address shortly.

The New Customer Acquisition Goal

Google also introduced a smart bidding goal specifically geared towards acquiring new customers. Users can bid higher for new customers, and this is called New Customer Value mode. In this case, users can continue targeting existing customers, with new customers as a priority. The other option, New Customer Only mode, only shows ads to new customers. This provides an efficient and consistent way to source new customers automatically.

This feature comes under the Performance Max campaign type, a solution dedicated to automation. Another helpful tool is the budget bid strategy, automating campaign budget management.

How AI Improves Query Matching

AI is improving the relevance of search results thanks to its advanced query matching capabilities, which may improve ROI on ad spend. A few developments in recent years are discussed below.

Bidirectional Encoder Representations from Transformers (BERT)

The BERT update to Google’s query matching algorithm was released in 2018. It uses transformers, a type of natural language model that allows the programme to better understand context. It does so by processing words in a search phrase in relation to each other instead of individually and in order.

Updates to BERT furthered its capability to understand context. One update made it understand how prepositions could change the meaning of a word or phrase, allowing the provision of more relevant results.

For instance, if a user were to search ‘flights from Dublin to New York’, the model would understand that they are only looking for flights in that direction and would not serve results for flights from New York to Dublin.

Multitask Unified Model (MUM)

This update was launched in 2021 and far surpassed BERT’s capabilities, able to generate as well as understand language. It is also multimodal – it can understand the meaning associated with multiple media i.e., text and images. In practise, this means it understands the context of an image used alongside a piece of text; it understands why that image was chosen to be associated with the text. In addition, its ability to understand audio and video formats in context is continually developing.

BERT and its predecessors use different models for different tasks but the plan for MUM was to have one model that does everything relating to indexing, ranking, and retrieval.

It has been predicted that MUM will be used and developed between now and 2029 when quantum computers may be available to provide the ultimate semantic experience; Web 3.0 is now on our doorsteps.

The Impact of Advanced Natural Language Models on Paid Search

Language models such as ChatGPT and Bard are expected to be highly disruptive. In term of paid search, they could be beneficial in that they would serve up more relevant and accurate results, including ads. On the other hand, changes in user behaviour may lead to more clickless searches.

Advanced models like these synthesise information from various sources in order to provide the most complete answers to search queries. They far exceed the brief, sometimes one-line answers seen in current SERPs.

As a result of the increased value brought about by the information served by these NLP models, smart bidding may become even more critical, with the quality of ads (and their associated user experience) being more important than ever.

Another possibility is that content marketing will receive more priority than advertising. With ChatGPT and Bard enhancing the user experience to the extent that it might, genuinely providing value may become the only way to get users into the funnel.

The same applies with the MUM model. The model operates in 75 languages. An important step in gaining more authority will be translating content into the most relevant languages depending on which markets the business is targeting. (The algorithm will actually tag content based on its global utility.)

In the future, valuable content may simply out-rank any advertising material, no matter how well optimised it is based on today’s standards. In such a case, content will be no longer be ‘king’ – it will be the almighty. In addition, content with multiple formats will also perform better; text with images may not be enough.

How Will Advanced NLP Models Affect SEO Strategy?

With the Multitask Unified Model being able to understand complex language, there may be less focus on keywords. Google may simply no longer need to use keywords as its main method for interpreting information and its relevance. As a result, SEO may have a much less significant role (and perhaps one day, it will become obsolete).

Marketers that create SEO content primarily for the algorithm will need to change their strategies, prioritising the user. They should be doing this anyway but these changes will make it a necessity.

Other Uses of AI in Paid Search

Responsive Search Ads

Responsive search ads present variations on the ad copy for each user and test different combinations of headlines and descriptions for optimal performance. This allows businesses to effectively reach their target audiences by presenting them with assets they are most receptive to, thus increasing conversions and return on investment.

Advertisers can upload multiple versions of headlines and descriptions, and Google optimises them in real-time. The number of combinations possible provide a thorough basis for testing and optimising results and, as with any Machine Learning model, the results will improve with time. Many hours are saved that would have been spent split testing and analysing outcomes.

AI Ad Content Generation

While AI content generators can help create marketing assets through their templates, Google Ads now contains a tool for generating ad assets for responsive search ads. The AI not only optimises the combinations of assets that the user has provided, but it also creates new versions. Early adopters have been seeing increased conversions when using this technique.

Predicting Ad Performance

Google can now predict the click through rate of an ad for a specific keyword. Its prediction comes in the form of a status: above average, average, or below average. If a below average status is returned, the business will want to revisit how closely the ad copy relates to the focus keyword – or whether the keyword itself is useful.

In addition, it predicts how the quality score is impacted, a metric indicating how the ad quality compares to that of other advertisers. (Ad quality is a measure of user experience when encountering ads.) The quality score relates to individual keywords and is measured on a scale from 1-10.

As well as the expected CTR, other key factors considered when predicting the quality score are ad relevance and landing page experience (defined as the relevance and usefulness of the landing page).

In Summary

AI has brought about many valuable developments in paid search advertising. Smart bidding has streamlined the process of managing large and complex campaigns, increasing conversions at the same time. Google can also optimise campaigns through responsive search ads, the generation of ad assets, and its predictive capabilities.

Aside from these use-cases, the bigger picture is that the whole landscape of paid search is likely to change dramatically. Advanced query matching based on natural language models is going to change user behaviour, and this is already happening judging by the increase of conversational search queries. As these models advance, smart bidding may increase in importance; on the other hand, paid search may become less and less relevant as utility will be the primary factor determining ranking.

Marketers will need to ramp up the quality of their content and provide a variety of formats and languages in order for algorithms to consider it of the highest authority. SEO may also become less important, with AI models able to understand entire pages without needing to look for specific keywords.

All in all, the semantic web is on its way and marketers need to start preparing now.

To discover how we’re helping businesses worldwide develop leading marketing strategies, contact us – we would be delighted to assist.

report cover

Download the full report

Would you also like to receive our CEO's market leading marketing newsletter?(Required)
This field is for validation purposes and should be left unchanged.