We didn’t exactly need a crystal ball to see fully search-integrated Gemini in Google’s future. But in late 2024, we got the first bit of hard evidence that the tech giants are actively working on the project.
Just another Google feature? Not exactly. Imagine AI Overviews were served without the traditional blue link listings, and only webpages cited by the overview gain clicks from searches. That’s the search format that ‘AI Mode’ presents — and wide take-up could have a drastic impact on your web traffic.
In this post, we cover everything we currently know about Google AI Mode, including capabilities, appearance, and projected impact.
Update (24.03.2025) - AI Mode performance insights from beta test
Members of the SEO community who are Google One Premium subscribers have been putting the current iteration of AI Mode through its paces - revealing the following key insights:
1. It assumes most queries have some element of local intent
AI Mode outputs frequently have locational context - even when the query is in no way related to proximity.
Whether this is by design or something Google will temper once the beta use data starts rolling in is unclear.
2. Searches for certain sites bypass AI Mode
We mentioned in a previous update that Google will present the usual blue links if it doesn't feel confident that AI Mode can effectively satisfy a query.
But Search Engine Land's tests have revealed that searches for the following sites also bypass AI Mode:
- DuckDuckGo
- Gmail
- CNN
- YouTube
- Wikipedia
- Semrush
This isn't an exhaustive list - just what's been reported.
3. Commercial and informational queries trigger the longest outputs
AI Mode is wordy-est when responding to commercial or informational keywords.
Not only are outputs for these intents longer in terms of word count (often upward of 500 words); they also tend to have the most citations.
4. Page 2 results are seeing improved visibility
When AI Mode is bypassed, the SERPs presented sometimes surface URLs that would usually appear on page 2 or as part of a featured snippet.
5. URL thumbnails are quite small but essential
When they appear, URL thumbnails are cropped to 82x82 px.
85% of cited URLs appear with a thumbnail, reducing the likelihood of clicks for the 15% that appear without an image.
6. Certain brands are dominating
A couple of US brands are already dominating responses to commercial search queries in AI Mode, with BestBuy leading the pack.
7. In-output citations often outnumber sidebar citations
For example, the query 'lightweight hiking boots reviews' garnered 21 in-output citations presented via anchor text - but this same output offered just 3 sidebar citations.
8. Referenced text is highlighted on source pages
When a user clicks through a citation in the AI Mode's output, the information the AI used to formulate the output will often be highlighted on the source page.
This is both handy for users and helpful for SEO, as these highlighted sections can be analysed to shine a light on what Google is looking for in a citable passage.
9. AI Mode for mobile shows fewer citations
On average, AI Mode outputs on mobile devices contain roughly 50% fewer citations than desktop outputs.
10. Informational queries produce the fewest citations
AI Mode currently takes a light touch to citations in response to purely informational queries.
Simple informational intent keywords generate as few as 4 citations - the lowest recorded citation count during the test.
11. Blue links may be surfaced in place of traditional SERP features
For example, the report states that online travel agencies are seeing enhanced visibility due to 'book flights' queries no longer triggering the Google Flights SERP feature.
It's unclear at this stage if the same is true for other query types and SERP features.
12. Historical authority is crucial
Reputation is a big deal to AI Mode. High authority websites are frequently cited in outputs.
13. Outputs show off Google's real-time & locational context capabilities
The AI search market is becoming increasingly saturated. Google is hoping that their access to real-time and locational information helps their product stand out in the crowd.
This may be why AI Mode currently adds specific geographic and temporal contextual statements even where not necessarily required – often it appears to be more of a flex than an effort to offer additional value.
14. Super local outputs for brand queries
Even queries related to large, non-consumer-facing brands, such as NASA, triggered hyperlocal outputs.
This is particularly interesting because Google appears to be removing local impact of larger national and international brands in traditional SERPs - in favour of consumer-facing outfits with a focused service area.
15. AI Mode shows occasional errors
Every now and again, AI Mode fails to produce and output, giving way to an error message that reads, 'something went wrong with this response'.
16. Outputs may currently include irrelevant citations
During these tests, it was discovered that AI Mode sometimes cites a source completely irrelevant to the search query.
This shows that Google still has some teething issues to work through - and a wider release may not be scheduled until mid-to-late 2025.
Update (18.03.2025) - Google confirms AI Mode ad tests
In conversation with ADWEEK, Google has confirmed that they will 'explore bringing ads into' AI Mode, leveraging insights from their experiments with ads in AI Overviews as a basis.
No timeframe has ben announced, but if AI Overviews is anything to go off, ad tests will likely commence following the current beta testing.
Update (10.03.2025) - Google launches AI Mode as a Search Labs experiment
Google have officially announced the next step in AI Mode development - making the feature available for Google ONE AI Premium users opted in to Search Labs experiments.
In Google's post, we also got more insight into how AI Mode will function when an AI-generated response may not sufficiently or responsibly address a query:
'AI Mode is rooted in our core quality and ranking systems, and we're also using novel approaches with the model's reasonng capabilities to improve factuality. We aim to show an AI-powered reponse as much as possible, but in cases where we don't have high confidence in helpfulness and quality, the reponse will be a set of web search results.'
This is precisely the process we possited in our post AI, answer engines and search: What you need to know. This way, Google nods to the importance and value of traditional web results, while pushing forward with their AI product.
What is AI Mode for Google Search?
In a nutshell, AI Mode will be Google’s version of ChatGPT Search.
AI Mode will begin as an optional search format that users can activate to trade Google’s traditional SERPs for an AI chatbot.
This mode will allow deeper discussion and exploration by way of conversational responses, the option to ask follow-up questions, and multimodal capabilities, i.e., voice and image search.
Like competitors, it will cobble together relevant information from a database (Google’s index) and rephrase it in a generated response.
Powered by a custom iteration of Gemini 2.0, it will reportedly have advanced reasoning capacities, helping it understand and effectively respond to highly complex queries.
Why publishers should be concerned about Google's AI Mode
We may be in the middle of an AI boom, but beyond professional contexts, AI is yet to really take centre stage and become normalised among the general public — Google AI search could do just that.
ChatGPT and similar tools are already enticing people to drop traditional search for a more flexible and conversational alternative, removing classic SERPs (and all the clicks they broker) from the equation.
But considering Google’s vast user base, broad implementation of full Gemini search would set this migration into hyperdrive. We could see a sizeable and sudden influx of users making the switch, drastically impacting website performance as SERPs are left behind.
Where will AI Mode be available?
Initial reports suggested Google’s AI Mode would hit Android mobile devices first. However, it has since been confirmed that there is a desktop version in the works.
How was it revealed that Google is working on AI Mode?
At TDMP, we’ve tracked Google’s integration of AI into Search for years, analysing its implications, predicting developments, and reporting on progress at several key stages:
- Google Bard, ChatGPT, Microsoft Bing: How AI is changing SEO – A deep dive into early AI-driven search shifts.
- Google to introduce generative AI responses in SERPs – Our early coverage of Google’s move toward AI-powered results.
- Preparing for Google SGE – full breakdown – Insights on how Search Generative Experience (SGE) would inform search strategies.
- Google launches AI Overviews: Key features and updates – Tracking Google’s transition from SGE to AI Overviews and its immediate impact.
We’ve been anticipating the arrival of Google Gemini Search for some time, first bringing it to your attention in our May 2024 article ‘What Google’s March update reveals about the future of SEO’.
However, the most concrete evidence of an AI Mode nearing completion came from 9to5Google. They did so by “decompiling” the files (APKs) that make up the latest version of the Google Search app for Android.
This process reveals code that serves as clues as to what Google is working on — and even what features larger projects could have upon release, but bear in mind that nothing is certain.
As 9to5Google explain, ‘Google may… not ever ship these features’, but it’s an intriguing glimpse into what could be.
Google is currently testing AI Mode internally
Since the initial revelation, a leaked internal Google email asking employees to test AI Mode has shed more light on the project.
In the email, Google describes AI Mode as ‘Search intelligently researching for you — organising information into easy-to-digest breakdowns with links to explore content across the web.’
They also mention that this new search format is designed to handle exploratory questions that are currently ‘not well served’ by typical search results. Here are some query examples given in the email:
- ‘How many boxes of spaghetti should I buy to feed 6 adults and 10 children, and have enough for seconds’
- ‘Compare wool, down, and synthetic jackets in terms of insulation, water resistance, and durability’
- ‘What do I need to get started with aquascaping?’
- [Follow-up]: ‘What are some nearby stores to buy supplies?’
What does the code reveal about AI Mode for Google Search?
The code referring to AI Mode (given as ‘aim’ in the code) can be traced back to October 2024 but has continued to appear in more recent iterations of Google’s app.
Reportedly, below are some examples of the code in question:
<string name=”googleapp_aim_mic_button_content_description”>Microphone</string>
<string name=”googleapp_aim_listening_text”>Listening…</string>
<string name=”googleapp_aim_photo_button_content_description”>Take a photo</string>
<string name=”googleapp_aim_gallery_button_content_description”>Add image from gallery</string>
<string name=”googleapp_aim_followup_hint”>Ask a follow up…</string>
<string name=”googleapp_aim_interrupt_button_content_description”>Stop loading</string>
Note that there is a reference to a ‘mic button’ for voice search and ‘photo’/’gallery’ buttons for image-based additions.
There’s also an ‘Ask a follow-up’ option — as is expected of an AI search facility.
What does Google AI Mode look like?
Below is a screenshot of the AI Mode UI for desktop; it is the only UI image available at this time.
Source: Searchengineland
Based on this screenshot, it appears that users type queries in the normal search bar and then activate AI Mode using a new filter that appears next to the usual options, i.e., ‘Images’, ‘Shopping’, ‘News’, etc.
By keeping the standard search interface intact while offering AI Mode as an optional feature, Google could limit the inevitable negative impact on web traffic, ensuring that users continue to interact with organic search results. This would also help protect the viability of traditional Search Ads and Display Ads. As we mentioned in our guide to SGE, if these ads stopped generating traffic, Google would lose a significant portion of their revenue.
AI Mode itself looks like a large AI Overview sans traditional SERPs — instead offering a card to the right that contains links to explore the web.
As you can see at the bottom of the screenshot, there’s a field available to ask follow-up questions.
Although no mobile device UI screenshots have surfaced yet, Android Authority managed to activate the mobile UI that contains the AI Mode button. The button in question features a magnifying glass icon with a rainbow border on the top left-hand side of the UI pictured below:
Source: Androidauthority
It didn’t work when tapped, as the feature is still in development, but it’s interesting to see how access looks — at least at this early stage.
The Android Authority team also spotted a button for AI Mode in the ‘Shortcuts’ menu in the Google Search widget:
Will there be ads in AI Mode?
As stated in the update added to the beginning of this post - Google have confirmed that they will be testing ads in AI Mode, using their learnings from ads in AI Overviews as a foundation to work from.
Google struggled to find a sweet spot for sponsored content in AI Overviews due to the delicate matter of balancing engagement between the overview, paid content, and organic listings. But AI Mode simplifies things by removing this last from the equation.
It’s possible they stick with the format currently seen in US AI Overviews (ads placed just below the overview), but Google could also explore more creative avenues, such as in-chat native advertising – which we previously discussed as a possibility for ChatGPT ads when they launch. Read our October 2024 AI & digital marketing roundup for more information.
How can businesses prepare for Google Search AI Mode
Preparing for the public release of Google’s AI Mode should centre on building visibility in AI search responses, chiefly — AI Overviews and the Gemini app.
These are currently the most relevant windows into how AI Mode will function and the type of sources it will cite.
We know that - for most industries - there’s a growing overlap between AIO citations and webpages cited in Google’s traditional SERPs, so your SEO strategy will play a major role in readying your brand for AI Mode.
However, before implementing targeted, AI-specific optimisations, businesses must first measure their current visibility in AI Overviews and Gemini. This means analysing how often and in what context their website is cited, identifying gaps, and tracking patterns over time.
A data-driven approach ensures that every optimisation is informed by real insights — maximising the chances of securing a strong presence in AI-driven search results.
Prepare for AI Mode with TDMP
At TDMP, we are developing advanced methods to track attribution in AI-generated outputs, helping clients understand where and how they appear in AI-driven search results. With these insights, we’re able to refine digital optimisation strategies to strengthen visibility in these increasingly popular spaces.
Get in touch to see how we can help drive your success as search behaviours shift.