Google I/O 2023: Key Points and Thoughts for the Search Industry
It's been a couple of days since Google I/O and I've reflected a little on the key themes and points. Yesterday, I shared a bunch of thoughts, ideas and notes with the team at Aira. Below is an edited version of what I shared with them.
If you're an Aira client reading this (hi!) and would like to chat more about this, drop your account lead an email and they'd be happy to put you in touch with me. If you're a brand who are looking to stay ahead of what comes next and don't work with us yet, you can email me here to arrange a chat.
Some thoughts on the Search section of Google I/O and the potential impact on clients
Everything in this section relates to this 13 minute segment from Cathy Edwards. If you work in search and haven't watched this yet, I'd highly recommend that you do. Here are a few themes and ideas that I took from Cathy's keynote.
A clear expectation that people will change how they search
The clear highlight here was the introduction of generative AI at the top of search results. The interesting part for me was that there were more than a few references to changing how people search. Here are some examples:
- "Things that you may never have thought to ask search for before."
- "Turn to Google for things that you never thought you could."
- "Unlock new ways to search."
Google are recognising that conversational AI is different to search as we know it right now - they're trying to bring the two products together into one interface. Personally, I'm not convinced that they are compatible in the short to medium term. Is it too much of a behavioural switch? I can see behaviour changing, but very slowly.
The current design also looks a bit, clunky... particularly with the links to sources in the top right corner:
I'd expect this to change over time and be refined more.
Trusting generative AI content
One quote that stuck out to me was "People will always value the opinions of other people" - context was that no matter what AI can do, people will eventually want to know that recommendations/information etc comes from a person. I'm going to agree to a certain extent, but there are also a bunch of queries where people just want facts that aren't debatable or about opinion. So a bunch of queries are at risk of never sending traffic to websites here imo - more on this below.
There is a legitimate question on whether users will even notice that content is from AI as opposed to a human. Assuming that Google keeps marking this content as such, which they strongly implied that they would across multiple formats, I think this is less of a concern.
Thinking beyond the new feature above for a second, where it gets really interesting for me is a scenario where a website publishes AI generated content (not human at all) and that content actually does answer the query of a user who arrives from search. Should Google care that the content wasn't written by a human if the user is happy? I mean, they probably shouldn't if the user is happy, but I suspect that they will care - publicly at least. I don't know what I would do if I were Google here, thankfully, that's not my job! But it's an interesting one to think about because I think that they'll need to balance their desire to satisfy users against the fact that they can't appear to let the floodgates to AI generated content on the web open.
Google also talked about identifying "synthetic" content (specifically, images) via watermarking and metadata. Whilst focused on images, you have to figure that they are looking at all content here eventually as a way to identify AI generated content.
Interesting way to describe it too - I'm predicting that we'll see "synthetic content" added to Google Webmaster/Spam guidelines within 6-12 months.
The impact on search as a channel
Organic search is a risk for sure to some industries (a few noted below). But I still think it's too early to know what the actual impact will be - we all need to play with Bard/Generative AI search to see how it may change things. For me, I still think that AI has a huge trust issue around informational/comparative content. Facts and statistics are easy to get right, but so much of search is people going on a journey and at some point, they want to be told what to think or do by a human. I'm not sure how AI bridges that gap in organic search.
I'd be scared and thinking about how to adapt if I'm any of the following:
- A template website of any kind e.g. legal documents, HR templates, health and safety docs etc.
- StackOverflow/tech support forums.
- Pinterest or any stock image website (more on why later)
- Educational sites that don't have a USP or brand that users buy into.
In regards to paid search, it's not going anywhere and needs to grow to keep shareholders happy. I'd expect it to get more expensive at least in the medium term if generative AI is integrated more into SERPs before Google monetises it - I wonder if they will take a short term hit and show fewer ads, but CPCs will increase to offset things. I wouldn't be surprised if Google reduces the number of Ads at the top of SERPs and drives the price up.
Speaking of revenue for Google, AI as a revenue stream for Google is interesting to think about - there were lots of partners being talked about in the Cloud section of I/O in particular. I suspect that the revenue potential from API access is a drop in the ocean for Google. So I think that partners will be allowed to use AI models by Google in exchange for sending data back in the other direction to refine and help Google learn. PaLM 2 API access are examples of this I think.
I also can't help but think about the cost of integrating AI into each search. Chat GPT costs are high - Google may be able to benefit from their scale here, but obviously has a lot more searches than Microsoft. Does this scale? You'd have to assume that it does because that's what tech does, but if they also want to get more complex with what AI can do, that requires more processing power, so scale savings may get cancelled out.
Robin also pointed out to me that this is actually part of Satya Nadella's game plan - Google taking a hit on margin on searches hurts them more than Microsoft taking the same hit. Whether this actually happens is another matter, but I think that it's definitely a challenge that Google will be addressing.
What brands should be thinking about
Firstly, don't panic. I've seen a lot of "hot takes" on LinkedIn and Twitter over the last two days and frankly, I think far too many marketers try to be the first to react to these kinds of events, rather than taking a step back and actually thinking about things. So my first piece of advice is not to be one of them.
Secondly, there is clear disruption happening and with disruption comes chaos. Fortunately, chaos and disruption lead to creativity and innovation. So once any initial panic has subsided, I'd be looking at this as an opportunity. SEO in particular has been dead for about 15 years now, we're used to it.
To make things tangible, I do think that brands need to adapt in a few ways. I'm thinking more about this at the moment, but a few initial thoughts come to mind:
- If you've lost organic impressions or traffic already because of featured snippets/answer boxes pushing you further down the SERPs, I'd expect this to get worse and plan for it. Look at the queries that are most at risk and forecast what the downside will be if this gets worse. I suspect that the queries that are already featured snippet heavy are the ones that generative AI will chip away at even more.
- If you provide any kind of content that can be seen as single answer or factual, I'd expect to lose traffic. Again, we've seen this already from Google where they're able to provide a concrete, clear answer to a question within a SERP - meaning that the user doesn't need to click through to a website. If any of your content provides answers to a question that aren't opinion, controversial and literally a fact, then I'd plan for losing traffic.
- Content at the consideration and decision stages of the buyer's journey is about to get a whole lot more important. I think that companies who invest in brand, product and content at these stages of the buyer's journey are more likely to win, even if generative AI within SERPs takes off.
The impact on ecommerce brands
The keynote had some clear examples of commercial queries, one being buying a bike. Links to more products were very prominent within the generative AI box and is built upon their existing Shopping Graph. You have to bet that they have already talked about monetising this.
I did notice on these results that key attributes of a product or retailer stood out massively. Things like discounts, deals, delivery times and return policies all stood out as a point of differentiation. One example search result was "Here are some dresses with two day delivery" - retailers need to enable Google to understand these types of USPs/attributes that they have if they want to stand out.
I think that ecommerce brands need to truly think about what makes their service or product different to everyone else and what value they add to the buyer's journey and the problems that a buyer experiences at the start and middle of that journey.
For me, this isn't new - ecommerce brands who aren't thinking about this are already behind. The fundamentals are still the same, but those who don't address those fundamentals will be left behind with these types of changes from Google.
Changes to local results
There was a brief glimpse into how local results may change with an example comparing different restaurants that a user may want to visit. The result shows three different restaurants side by side and each one had different attributes being displayed.
Like retailers, I think local businesses are going to need to give Google this kind of info to stand a chance of being here in these results.
The impact on image search
Google shared that they have partnered with Adobe Firefly which allows you to use Bard to create images on the fly using Adobe images and technology. I feel like this may have an impact on image search and copying images. You'll basically be able to create unique images for whatever you want, so going to copy them from elsewhere isn't needed as much. I'd worry if I was Pinterest or any stock photo website.
They also talked about improving search results by allowing someone to click a button that shows them more information about the image.
Highlights and further reading
Before going into the highlights and cool new products/tech that was announced. An important piece of context should be kept in mind - the vast majority of Google's revenue still comes from Ads. Sure, their stock price jumped by 5% after Google I/O, but ultimately, their long term roadmap still leads towards selling more and more Ads. So I'd advise keeping that in mind when you think about these updates and where they may lead to.
Interesting product changes and innovations related to AI
- Help me write - expands on smart compose and will write an email reply for you, using details from the incoming email. You can then refine by making it longer/more formal/elaborate. Pretty nifty little feature and I think it will get used a lot. I'd be interested in how we could/should use it for prompting good replies to client emails.
- Immersive maps are being introduced for 15 major cities by the end of the year which allow you to preview a route before you take it. Pretty cool for walking or biking through cities. The visual demonstration was very impressive and it integrates the weather and air pollution, so you can see if it's likely to rain later when you start your walk.
- Magic editing is being introduced. Advanced photo editing - made me a bit sad tbh, it almost feels like too much editing e.g. changing poses and adding extensions of objects that aren't in shot or taking away clouds to make it look like a sunnier day than it was. Super cool, but I'm not sure that's what photos are about.
- Workspace new features:
- Help me write - within Docs, you can use it to help you write whatever content you want. I think this may have a big impact on templated document websites such as legal / HR templates / employment contracts etc. The legal industry generally is at risk here I think with the less customised stuff that they do.
- Help me organise - within Sheets, you can ask it to create a table for you with an abstract description and it will do it. Super cool.
- Help me visualise - within Slides, you can create images from text in slides to make things more visual. You can also ask it to add speaker notes that are based on your slides. This could have a useful side feature which is to get a "summary" of your slides and see what the takeaways may be.
- Getting prompts within Google Docs (called Sidekick)
- They've identified that sometimes, people need help with prompts, so they have developed "Sidekick" which sits at the side of a Doc or email. It helps people continue when they get stuck with writing. For example, you get a few paragraphs into an article, need some help, but don't know how. Sidekick reads the text and gives you some prompts which you can click on and get ideas from. It occurs to me that the same tech could also be used to help someone continue their journey in search too.
- Sidekick in Gmail - same kind of thing as above but the use case was to summarise email threads and even reference Docs/Sheets/Slides that are included in the thread. Could be mega useful for summarising long email threads with clients.
- Bard new features/integrations - it's clear that they are trying to connect Bard with more Google products. A few examples:
- Export responses directly into Gmail, Docs and Sheets.
- Export code to Collab and Replit.
- Upload an image to Bard and get captions generated to describe that image - my first thought was ALT text generation en masse would be a potential use case here.
- The presentation really focused on Bard as a standalone tool outside of search i.e. not part of Search itself (although that came later) and then a "Google it" button which allows you to go to Search if you'd like. But it feels like they aren't doubling down on Bard within Search alone - it's still going to be a standalone tool.
- From within Bard, you can bring in other Google Products e.g. Maps. So the question is - what else can they bring into Bard?
- They are already connecting with 3rd party apps e.g. Adobe / Redfin / Spotify / Indeed.
- Project Tailwind - scans Drive and uses the content as expertise. So you can essentially get content recommendations from AI based on your own tone/expertise/style etc.
- They are developing more tools for evaluating information and trust.
- About this image tool - will show when similar images first appeared and were indexed. All Google generated images will have metadata to help identify it. Publishers and creators can do the same and mark them in search as AI generated. You have to bet that they are trying stuff here with other content forms (video and copy) too.
- Universal audio translator is a really cool new tool that translates audio in realtime.
- Could open up education and learning to new territories without needing to have human translators. Videos/Podcasts etc. Only accessible to authorised partners (for now) to help prevent misuse and deep fake enablement.
- Adversarial testing is something they are developing which is designed to help find fake content/misinformation and highlight it. For example, helping dispel fake moon landings photos - the tech here is surely being used for general information/credibility too on the web too.