I have had access to Google's new Search Generative Experience (SGE) for 1 week now. Here are some of my observations from testing the generative AI across dozens of clients at Graphite 👇 I believe that this update presents significant traffic erosion potential for *some* types of queries and page types. - Some businesses will be heavily affected by this update, others will be minimally affected (depends on industry and current sources of SEO traffic). - Different page types present very different risk profiles for traffic loss. Page types at high risk: - Listicles: Google is essentially creating their own listicles now directly in search. The standard commodity listicle that has worked for so well for so long will be a thing of the past. - Category pages (browse): Similar to listicles, Google is now essentially creating category pages in search results. Category pages will be pushed farther down the SERP and receive far less clicks. - Top of funnel educational content: Anything that can be directly answered by the AI is at risk (ex: “what is”, simple “how to”, etc). Page types at medium risk: - YMYL ("Your Money or Your Life") topics: Interestingly, Google is showing SGE for many health related topics (ex: SGE shows "diabetes symptoms" and "tips to reduce blood pressure" / is absent from COVID and RSV queries), but is NOT showing SGE for anything related to finance. Page types at low risk: - Individual inventory pages (ex: eComm products): I actually expect these to increase in traffic significantly due to Google now ranking PDP’s in SGE for queries that used to trigger PLP’s. - Breaking news: Hard and high risk to summarize with AI. - Anything where you actually need to DO something on a website and can't be satisfied through an answer (ex: stream a show). Google mentions brands directly in AI snapshots. - Brand building will be even more important as the AI snapshot often mentions brands by name for non-branded queries. SGE seems to bias towards ranking UGC and forum sites. - One of our client has a programmatic page type with thousands of pages that rank very well for long tail queries. Each of these pages has an equivalent page on the client's community forum subdomain. - SGE is ranking the community site more frequently in the AI carousel than the main site, despite the fact that the main site tends to rank significantly better in traditional organic results. Google doesn’t feature in-text links in AI snapshots. - SGE has linking carousels and will feature products/reviews, etc… but it doesn't link specific words in the AI snapshots (for now). SGE is sometimes wrong / makes up bad information. We all need to reconsider keyword intents, as certain keyword types will become more or less important in SGE. - Ex: “freelance social media marketer” and “hire social media marketer” show 2 completely different SGE responses, while the traditional organic results are almost identical. Bonus video attached to show live SERP examples!
Search with Generative AI does not make it look good for measuring top-of-funnel traffic accurately and will lower traffic perhaps on discovery pages. Generative AI can/will also help by ranking products or features by importance and/or provide the most important criteria in a table. It makes product comparison so much easier as well as you'll understand all the product criteria way better. It will become a decision making/driving engine and the bigger topic is what it means for marketing as a discipline. It will force the marketer/companies to make better products to rank higher, but does it give an unfair advantage to established players as it will be harder for new entrants/products to get to product-market fit? Marketers will have to optimize even more so, for the customer and to understand (behavioral) journeys (measurable or immeasureable) to optimize for generative AI. I think the bigger topic really is, that the way how it works or what it does needs to be open sourced so that transparency to the consumer is there. How/Why do you get presented with certain products, what drives it, how can you make the playing field level. Prepare for the influence of (dark) generative AI search in your marketing funnels.
Good insights! I would disagree slightly on category pages being labeled high risk, though. I could be wrong, but one thing not touched on here is the power of a strong brand. I’m not sure users will want to stop browsing on specific sites. The example you gave rings true for aggregate review sites for sure, but I don’t know if a curated PLP with different brands is the ideal UX/UI. I think we can still use good old brick and mortar for insights into consumer behavior. If someone wants to buy a blue shirt, they’ll go to the mall and check out different shops and make subconscious brand judgements along the way which influence their future actions on where to shop. If you put all the blue shirts that are in a shopping center into a single store front, no one would shop there. It would be way too overwhelming. A GSE-generated cat page for the kw “blue shirts” is the equivalent of a single store front with millions of blue shirts. I don’t think the world is ready for brand-agnostic shopping experiences.
For SaaS, would heavy product-led content and queries be the way to go? Hard for Google to provide in-depth AI answers for all software. Less traffic but possibly more conversions?
There are some really interesting developments being made in integrating vector databases with LLM's to facilitate search functionality. This is one article on the topic but there are also great talks by Redis: https://www.elastic.co/what-is/vector-search While it's hard to say what google will do, I imagine they will post-process their results to include reference to their ad clients at the bottom of the answers.
Is this what the average user wants...? I don't, I may be biased because I'm in the industry, but I truly dislike it. For shopping queries, yes, it looks far superior but overall I see less clicks for organic and PPC!
This is awesome; thanks for sharing! What types of queries did you test to see SGE's impact on breaking news?
Do you like it though? I think it kinda sucks. I'm hoping it goes to its own tab like Bing has done - too slow, results are weird at times - just clunky.
Thanks for the summary - I agree with most. SGE is still in its infancy and a lot will change until the official launch. I said from the beginning (announcement) that branding will be become more important and totally agree that listicles and comparisons (as least if done without any decent background added) will loose out, as this sort of content can be easy generated by AI. Not so sure about the “how to” and “what is” – some very simple content pieces might loose out, but when if comes to more complex explanation, I think there is a big chance, if high quality information is provided. Basically, it is what Google was aiming for and communicating for some time now.
Technical SEO & Content Strategist | User Advocate
11moI've been testing too and have some similar takeaways: - PDP optimization and presence in Google shopping will be more important - I think that users will at times be frustrated with the AI results that aren't clickable. Ex. if I search 'streaming services' the AI gives me a list and as a user I'm expecting to be able to click through to those websites, but can't requiring me to keep scrolling or perform a secondary search to learn more about each suggested provider. My prediction is Google will have to add links to reduce user frustration - The AI article carousel is confusing on mobile (tested in the Google app). On my device it wasn't intuitive it even was a scrollable carousel due to no left & right arrow showing - Sometimes the AI generated results aren't relevant to the user search intent. I think Google will have to work to improve this before a full rollout, as they earned their reputation as the best search engine due the quality of their organic results