AI's Impact on Search and Marketing

AI: The next evolution of marketing

As technology innovates, marketers are provided with new ways to grow and evolve. Today ChatGPT has become the “next big thing” with jaw dropping rates of usage – gaining over 1 million users within days of its release.

One of the best ways to think about AI technology like ChatGPT is in terms of how the internet has impacted travel agents. As the internet evolved it became easier for individuals to book their own travel destinations and itineraries. Today one can do it as easily as using a search engine. However, despite third-party booking sites and consumers' access to price comparisons, the travel agency is far from dead. Rather, what was once a transactional interaction has turned into an experience-based one. From Disney vacations to cruises to multi-destination trips, good travel agents continue to offer tremendous value.

For marketers, the expectation remains consistent. The good ones will continue to thrive and find new ways to move themselves, their clients, and the industries they serve forward.

How will AI change search engines?

In 2013, Google released the 'Hummingbird' update, which began their transition from a lexical (keyword-based) search engine to a semantic (intent-based) search engine. Today, we experience a mix of both when searching on Google. For example, before the update, a search for 'weather' may have yielded results about trending weather news or a definition of the word 'weather.' Now, Google understands that the intent is most likely for the local weather forecast. This integration of natural language search (NLS) made it easier for searchers to find answers to their intended search, not just the query.

Building from natural language search, conversational search is the use of natural-sounding sentences or phrases to give the best results for a searcher’s query intent. This evolution of search is essential when thinking how virtual assistants—such as Alexa, for example—understand voice queries. Virtual assistants must be able to 'understand' what they are asked in the context of normal speech and then how they will provide correct answers to those questions.

As Google continues to evolve, such as with the introduction of 'RankBrain' in 2015, they have leveraged the use of AI to determine the most relevant results to a user’s queries.

That’s to say, AI has already had profound impacts on search results; however, the introduction of highly advanced chatbots has opened new possibilities. More specifically, in relation to the integration of conversational search and the results it shows the user.

While new AI tech is exciting and ChatGPT is making headlines, the only surprise is that the technology is being unveiled—not that it exists. As it has been reported that Microsoft intends to integrate OpenAI’s ChatGPT into Bing later this year, we can expect Google to launch their own version (LaMDA) in the not-so-distant future.

The key component will be how both major players (and others who will enter the space) provide additional inputs into these platforms. Specifically, a major limitation for ChatGPT now is that it can only provide answers on the data it’s been trained on—which is a small subset in the grand scheme of all the knowledge available to users (as the internet claims to hold). Both Bing and Google can solve this problem by integrating their chatbot solutions with the robust knowledge bases each have been building for years.

A significant question to anyone who has used ChatGPT is if chatbots will make search engines obsolete. While chatbots provide searchers with a new level of convenience and ChatGPT continues to show its ‘intelligence’ and ability to mimic human speech patterns, there are still obvious issues:

The outputs can provide incorrect answers to queries. This can be true for pages ranked by Google now, but users can see a source and determine its validity based on the domain (though this is not foolproof, many users can be ‘duped’ by spam sites.
- Expanding on this point, chatbots do not cite sources. Google has also been trying to counter this through its E-E-A-T guidelines which focus on 'experience,' 'expertise,' 'authority,' and 'trust' in how it determines the value of and ranks a website.
These technologies aren’t created magnanimously. Ultimately, they will be monetized. While search engines have solved that model, it doesn’t appear that chatbots have—the natural assumption being they will have a pay-to-play model for users, rather than for corporations.
Chatbots need to be fed information to be successful. If chatbots replace search, then businesses will be less inclined to provide content that won’t drive to their site or drive a sale. Without new content, chatbots may struggle to maintain relevance, thus forcing momentum to swing back.

How marketers should and should not use AI tools

Will AI take your job? No. In its current state and near-term outlook there is no reason to believe that chatbots will replace the job of marketers, including SEOs and copywriters. Though, it may necessitate leveling-up skills and bringing uniqueness and value to the products those in the industry create.

AI tools can be great facilitators and efficiency-makers for marketers, but they can’t do everything that humans do and, most importantly, not as well. To truly target and connect with an audience, a human perspective is vital to ensure elements like storytelling, compassion, and empathy are included. AI can’t ‘do’ marketing and it can’t understand the psyche of a consumer and marry it to a value prop.

Content-Generating AI Tools

Content-generating AI tools require limited text inputs to develop high-quality outputs that have incredible efficiency. They can be used for anything from generating product descriptions to writing full blog posts. The image below displays an example product description that was created using Copy.ai.

While these tools can produce well-written content, it is not recommended to utilize AI-generated content as published web content. Search engines, particularly Google, can and will continue to evolve their ability to detect AI-written content and will remove it from the search engine index as it directly violates their spam policies.

Although it is not recommended to use auto-generated content as web content, a marketer could take the output from an AI tool and add valuable human input (input that helps customers and users) to enhance what was written by an AI platform. It is important to add valuable human input as AI-powered tools lack legitimate expertise and understanding of a customer's needs. After all, it is the job of marketers to understand what drives a consumer’s need and then determine strategies to solve that need.

Chatbot AI Tools

Chatbot AI tools (ChatGPT or LaMDA) that have been trained on a vast amount of data are very powerful—they enable users to ask questions and receive high-quality textual responses. Although these chatbots can generate content in a comparable way to content-generating AI tools, they can lack uniqueness. Because of this, it’s important to remember that the primary function of chatbot AI tools is to hold a conversation with the user.

Marketers can utilize this technology in numerous ways, from asking the chatbot to write regular expressions for improved reporting to creating a blog article. To reiterate, however, using outputs from AI tools will directly violate Google’s spam policy and will negatively impact your, or your client’s, organic performance. Equally important, utilizing non-human tools can create impersonal relationships and hurt consumer loyalty.

When using these tools, rather than asking the chatbot to write a blog article, a marketer should prompt the tool to write an outline on a topic to speed up the content creation process.

The use cases for chatbot AI tools seem nearly endless and one of the most impactful uses for marketers is leveraging the tool’s knowledge to quickly learn about a business, clients, and/or industries. When consultants transition onto a new account, there’s always a buffer period that is required for them to truly understand a client’s business, industry, customers, and products. The images below display a conversation with ChatGPT that has rapidly decreased the amount of time it would normally take for a marketer to understand the client and their needs.

When using AI tools, it’s important to consider how the AI was trained—as the results produced from an AI are only as good as the data they are trained on. While the results may seem highly sophisticated, each one contains inaccuracies that a true subject matter expert could identify. For example, the first image refers to coatings as ‘thin films,’ however, a subject matter expert would consider this to be an inaccurate explanation of what a coating is.

In the image above, ChatGPT was asked, 'What are the leading coating companies in the US?' It did provide the user with useful information but contained a few flaws, exposing the inaccuracy of the output. For example, it identified 'Sherwin-Williams' and 'The Sherwin-Williams Company' as two separate companies. Additionally, it identified 'Valspar' as a still-in-business company, even though it was acquired by Sherwin-Williams® in 2017. Although Valspar was acquired by Sherwin-Williams®, it is still a paint (not coating) brand that can be purchased. These distinct nuances were not understood by ChatGPT.

AI-Powered Data Analysis Tools

Up to this point, this article has focused on AI tools that utilize deep learning to understand unstructured data (images, videos, voice, and natural language). But as marketers, we are consistently working with structured data (i.e., data in tabular form)—where AI-powered data analysis tools come into play. Like the previously discussed technologies, there are several data analysis tools that operate and function in various capacities.

An example of an AI-powered tool includes Google’s Entity Reconciliation API, which allows users to reconcile several datasets using semantic clustering and deduplication methods. Marketers can use a tool like this to merge their clients’ data to quickly discover valuable insights. This means that if a client’s web data and revenue data come from two different sources, the tool can help consolidate the data—a task that was previously done manually.

It’s crucial that marketers understand that while these tools can create efficiencies by developing reports faster and easier, they cannot necessarily derive insights that will drive performance. There are a few AI-powered analysis tools that promise to deliver ‘insights’, however, their inability to connect the facts with the business at hand makes their ‘insights’ surface level and very different from the deep insights that a marketer should bring to the table.

What’s next for marketers

While AI is poised to be quite disruptive in several industries, it’s vital that marketers understand that these technologies are not replacements for the invaluable work that they themselves do. Rather, AI-powered tools should be used in clever ways to create efficiencies and enhance deliverables. Similar to how the internet provided travel agents with an opportunity to offer customers tremendous value, marketers can leverage AI tools to offer their clients expanded benefits.

AI is trending across multiple platforms now and this technology is just getting started. Considering that transformers, an AI architecture that accelerated the possibilities for Natural Language Processing (NLP), were invented just 5 years ago and new AI architectures are developed every few years, the future for these technologies in marketing are certain.