<
How is Google using AI in search, what’s relevant for business and how will this affect business marketing?
There are currently two new Google search features gaining attention: AI Overview and AI Mode. Whilst we haven’t yet seen the full role out of Google’s AI Mode in Google search in Australia yet, AI Overviews are now common in search results with some reports indicating they appear in as many as 55% of searches. AI Mode in Australia is probably not far off so, if you haven’t already, it’s time to address how this will affect small business websites and digital marketing. In this article I’ll give an overview of AI in Google Search whilst considering the implications for business marketing.
- What is AI Overview in Google Search?
- What is AI Mode in Google Search?
- What is a large Language Model and what is Google Gemini?
- Where does the Google Gemini Large Language Model (LLM) data come from?
- Google Search AI Will Act Independently in Future
- Search Live
- AI Principles
- AI Data in Google Search Console
- AI Overview and AI Mode: Implications and Opportunities for Small Business
- How to build AI Confidence in web pages
- What is it all for?
What is AI Overview in Google Search?
We’ve had AI Overviews in Australia since October 2024 but it has certainly become more prominent recently in Google search results, with summary information and cited sources, including links. ‘Web resources that support the information’ and ‘relevant webpages’ sic. Interestingly, depending on the device I’m seeing AI Overview above other results, even ads.
The sort of searches that trigger AI Overviews are more complex searches and I’ve noticed, those that are question based definitely trigger the AI Overview more often. I’ve noticed if a search is updated after the initial response, the AI overview is triggered also, if it wasn’t before. To immediately trigger an AI Overview use words like ‘How to’ ‘What is’ and ‘How do I’ in your search query. LLM’s are used where a broad overview of a topic is useful.
In use, asking questions as search queries shows differently across desktop or mobile. On mobile for example you might get four sponsored ads appearing above the AI overview result whilst on desktop AI overview result and on desktop below.
I have noticed that currently featured snippets beat AI. But I strongly suspect AI Overview and AI Mode will replace those.
What is AI Mode in Google Search?
At the time of writing AI Mode has not yet been rolled out in Australia. When AI Mode IS rolled out in Australia a click on an AI overview is going to transition the search into full AI Mode.
AI Mode answers questions. It does this by breaking the question into sub topics that are searched simultaneously to provide a more comprehensive answer than our previous search has. Google calls this the query fan-out technique. AI Mode results provide an answer with the option to go deeper, with links to resources. If a single link is returned, that shows high confidence in the response.
Deep Search is the next step, with hundreds of searches generating the response going beyond the initial AI Mode query and using reasoning (!) creating ‘expert level fully cited reports‘ in seconds. This is mind blowing. You can apparently also add your own sources to Deep Research
Google notes that AI history and app data can be used if this is switched on, to continue a search where it has been left off. Your history can be deleted but even after deletion can hang around for 24 hours.
What is a large language model (LLM) and what is Google Gemini?
While I’m sure most of us are across what an LLM is let’s recap. LLM’s are basically massive datasets of human language (text and code) that have carried out deep learning using machine learning to find patterns in structure. They are trained by ‘fine-tuning’ to provide better responses. Google’s Gemini is an LLM that was announced to the world back in December 2023 (based on a previous version called BARD). Gemini 2.5 is the current version, a multi-modal LLM and generative AI used by Google in AI Mode search. Multi-model means not just text but image, sound and video.
Where does the Google Gemini Large Language Model (LLM) data come from?
According to Gemini 2.5 Pro, the source of the LLM data is the publicly available web; books; code and also internal Google datasets. Human feedback adds to this so, our search data becomes part of the generative AI experience.
Google Search AI will Act Independently in Future
Google Search AI will eventually be able to act independently on behalf of humans for what are called ‘agentic’ tasks. The opportunity to have AI book appointments for example. Google mentions that allowing access to personal context data will enhance these tasks. I assume access to email and calendars for example. Currently they are only noting google apps such as Gmail for personal context data.
Search Live
Remember that these days a search is not just text based. We have people using voice and images as well as text for the search input. Video is now also a search input. Multi-modal response is coming with Project Astra on the horizon currently only available for testing on Android devices. This is testing integrating live AI Mode with voice and video for realtime response and feedback to questions as they say turning search into a learning partner. This really is mind blowing.
AI Principles
If you are wondering about privacy now and how Google LLM and AI search data is used responsibly, you’re not the only one. All of the work Google is doing is guided by their AI Principles. More on that later.
AI Data in Google Search Console
So, now we come to the latest news that Google AI Overview and AI Mode data is now appearing in Google Search Console reports. It looks like this is generally counted the same way that previous search data has been and is grouped under the web search type.
AI Mode clicks count.
AI Overview assigns the same position to all links in the overview which I believe means multiple sources can rank number one.
Follow up questions after an initial search are considered new queries
AI Overview and AI Mode: Implications and Opportunities for Small Business
Let me throw some ideas out there…
To start with, our trusted organic search results are being pushed further and further down the page and AI Overviews and AI Mode will contribute to this further. We will be moving more towards answer engine optimisation (AEO) to build what I now call AI Confidence. Thankfully AI Search relies on similar factors for relevance such as expertise, authoritativeness and trustworthiness (E-A-T). I think it’s clear that the visibility of digital assets in multi-modal AI Search is going to be a goal for many. Web pages, images, video and audio content will need to be optimised for answer engines. Search engine marketing is surely changing to a more conversational version. People are talking about Generative Engine Optimisation (GEO) replacing search engine optimisation (SEO). The I don’t know if I would call it that – it’s still search.
The opportunity to build confidence with Google AI Mode
If AI Mode returns a set of web links instead of just one, I see that as an opportunity to do better and be the one helpful result that answers it all (eliminating the others from results). Remember, Google says one link equals confidence.
The impact of a click-less search response
We will definitely see the impact of link clicks lost due to our information populating Google’s response (yes our site can still be the source) rendering a click unnecessary. Google Search drives click-throughs to websites to start with and that leads to phone calls and form completions. Without a click through we may lose traffic but I do think this will depend on the query. I think any kind of intent to buy or exploration / evaluation phase searches will still land whether that is by organic or paid we will see. Zero-click searches are certainly a challenge.
Meta Pixel Audiences and website visits
The Meta Pixel is currently used by many advertisers to create audiences based on website visits. A lost click will reduce audience size. Again the intent of the search is probably most important here. Not all searches trigger an AI overview currently so perhaps this only impacts certain paid ad strategies around remarketing / retargeting that doesn’t involve a direct sales approach (such as content designed to educate rather than sell).
The Resurgence of Long Tail Keywords
AI and AI search is changing how people use search engines making long tail keyword phrases important again though conversational queries.
I wonder if all of those content heavy (awful) link building blog sites might also have a resurgence as they will be able to sell visibility again? I certainly hope Google will has pre-empted this. I think we will see a shift from link building to instead build AI confidence.
This can help with visibility during the exploration stage of the customer / buyer journey. I am assuming that means uncovering long lost pages buried deep in Google index. That is democratising really. Visibility for the little sites again perhaps based on hyper-relevance gained by answering niche questions relevant to what they do.
Optimisation for agentic search
Contextual personalisation for agentic tasks means that our emails including newsletters could form part of the search personalisation and become important. As one example does this mean we will see the rise of email optimisation for agentic search? Will Google AI mode eventually be able to complete our enquiry forms? Will our enquiry forms be bombarded with AI search triggered agentic completions? I’ll have to look into that.
AI, AI Search and Privacy
One of the biggest concerns for me is that now anyone can feed your shared data into AI apps and AI Search and it can become part of the LLM machine learning without your permission or knowledge. For example an accountant I was talking to recently mentioned that they are feeding customer data into Chat GPT to create reports in seconds. I’m sure it would be happening in Gemini too. For me, this shows a need for organisations to clearly communicate what they are doing with your data. Even your private data in someone else’s inbox could become part of the personal context and therefore competitive intelligence may be fed into LLM’s in future, dependent on their settings not yours. It probably already is as I believe opted out is not the default. Google AI features can use data from a users entire Google account history including Gmail, YouTube and Maps.
I believe the AI Search data is not identifiable and even when reviewed by a human, sensitive information is apparently redacted, but there are massive implications for intellectual property, competitive intelligence and what I think will be a diminishing ability to differentiate with unique knowledge. Remember Google’s current mission: “to organize the world’s information and make it universally accessible and useful” to deliver the most reliable and relevant information available.
The beautiful side of this is there will be a generation of people who are going to be starting faster uncovering that knowledge via AI Search. Sure, just because you have the recipe does not mean you’ll be a great chef but the amount of knowledge that is available will definitely speed things up.
How to build AI Confidence in Web Pages
So what what makes a web page a relevant web resource for an AI Search?
AI favours direct and concise answers to questions.
Consider how Gemini LLM fact checks content to find ways to build confidence.
Keep thinking about your website and web page structure and content hierarchy and how that builds topical authority.
Note that AI can identify trends and patterns. You’ll notice social media content has been surfacing in Google SERPS for a while now.
The path is now: Question > Response > Go deeper (follow up questions, helpful links). Matching the process of ask a question, follow up and explore may come into play for web pages that wish to build AI confidence and be visible via AI powered search responses.
A return to multiple search engines?
Another thing I am thinking about is agentic search and access to personal context content. This could drive LMM loyalty to a single search solution, the one that is most integrated with our other apps. This is a half formed idea but … maybe we’ll see a return to multiple search engines again like the old days when Google was one of many search engines in use.
What is it all for
We must always remember, the end result is to help the user. We are users. I personally am also responsible for leveraging this technology to assist business owners. When implementing anything online I am thinking about matching business goals with user goals. It’s all for people at the end of the day and building for people is going to get the results you want.
If you’re a client of mine you will be hearing from me personally, regarding what I’m doing to leverage AI search for your digital assets. Not everything needs to be live here *grin*.
Today, data is power. We are the fuel in the fire. I’m watching all of this and remembering the original Google search experience many years ago and the motto ‘Don’t be evil’.
by Alicia Laing, CREATIVE MODE