SEO in an AI world: first draft of a strategy
For the last month, I have been advocating the need to develop a new SEO approach to generative AI, and now that I have had a chance to experiment with Google’s SGE, I have more thoughts on what that strategy might look like.
My approach will continue to evolve as I experiment with SEO in AI, so this is a work in progress. Additionally, it is still very early in Google’s own journey into generative AI within search and how AI and search mix will continue to change. I have noticed that AI responses have stopped appearing on some queries it did before and then it comes back.
As I shared earlier this week, generative results appeared on most of the queries I conducted, but it’s highly possible that the ratio of generative to non-generative responses will go down as Google looks at how users experience this product. I don't expect Google to give up this project easily even if users are not loving the experience.
For those of you that recall Google Plus, it was never a resounding success or even something that users wanted; yet, they still embedded it all over the Google experience. They didn’t give up this project until 2019 - eight years after it launched.
Here’s my first draft of a new strategic approach to SEO in a generative first world.
From my experimentation, generative AI responses dominate the top of the funnel for most queries which means that all content that has been written or will be written for buyers early in a journey will be eaten by a generative response. For example, "hotels on Las Vegas strip" will be perfectly served by a generative response. "romantic experience at MGM Las Vegas" which is a mid-funnel query will likely be a click that a user will scroll to an organic listing.
Strategic Action: Focus on the mid-funnel rather than top of funnel.Query reporting will be obsolete as a metric for SEO success. Instead, you will need to focus on revenue from organic search as a KPI with clicks as a leading metric. For now, this might be difficult to accurately measure because Google clicks might not show a query at all in Google Search Console, but in whatever reporting tool you use, they should at least show up as referrals from Google.com.
Note, that in a query-less world, impressions and click-throughs will matter a lot less. Impressions will not be an accurate measure of visibility because users might not even scroll past the generative box if it answers the question. Clickthrough rates will be highly volatile as on precise queries it will be high while on broad queries it will be close to zero.
Strategic Action: Build out revenue reporting from SEO and stop focusing on keywords unless you want to see all of your reports blow up.Along the same lines of query reporting becoming obsolete, this also means that forecasting based on queries will also break. I have never been a big fan of beginning with queries as the basis for a forecast due to accuracy issues, but now they are even less helpful.
Instead, the best way to approach SEO forecasting is to use a calculation of the total addressable market (TAM) and then estimate an expected reach or penetration rate. This will be the subject of a full newsletter on its own on how I use this calculation, but essentially the TAM for a physical product won’t change just because search does.
Strategic Action: Create a forecasting framework that does not rely on predicted query volumes and clickthrough rates.For content creation ideas rather than using keywords from keyword tools, I am going to be using a lot more People Also Ask because these are already conversation starters and based on the longer phrasing it will be easier to identify mid-funnel ideas.
Within a generative box, there are additional follow up questions which right now I am seeing as different from People Also Ask, but it is entirely possible that these end up merging. Generative AI in search moves searches into a more personal and flexible experience rather than the rigid requirements of “head” and “shoulder” keywords that are relics of the early days of AOL search.
Strategic Action: Use a system to collect People Also Ask related queries to your product or service. There are a number of tools that do this and will be linked in my deeper essay on this topic.Everything about links and SEO are going to change in a world where search becomes a lot more personal. The often-abused concept of domain authority that had marketers chasing links from authority websites may not be necessary anymore. Generally, those links were most useful on the most competitive search terms, but if those competitive search terms are now the exclusive domain of generative responses, it is no longer worth obtaining those links just for SEO.
On the mid-funnel terms where the conversion rates will be higher, the competition won't be fierce enough that deep authority will be necessary.
Relevance and topical authority will become more important.
This is a lot easier and less expensive to do. You can prioritize cross-link relationships with related sites in a vertical rather than authoritative media sites. As an example, a local realtor won't need a link from Forbes to generate authority for a very specific town + property query, but if they have reciprocal links with other local realtors the algorithm will trust that they are truly local and specific.
Strategic Action: Look for sites in your “contextual neighborhood” that might give you a link to your website, and just ask.
Now is the time to pivot an SEO strategy before generative AI launches to all Google users. I think that this launch date is sooner than most people think. OpenAI is passing search queries over to Bing, and there is a substantial risk of Google losing market share if they don't get ahead of it. Even if this new generative product isn’t ready for primetime, Google can’t afford to hold on to it for too long.
I will keep sharing updates as I learn more, and thank you for reading!