Group items tagged

The average AI Overview is 169 words and 912 pixels long.Only 12.4% of the analyzed keywords display an AI Overview.A Featured Snippet is showing on 17.6% of the analyzed keywords.On average, AI Overviews appear alongside Featured Snippets in 7.4% of cases. For the Health niche, they show up together the most often (34.9% of queries).AI Overviews contain 7.2 links on average when expanded.33.4% of AI Overview links rank in that query's top 10 organic results.46.5% of the URLs included in AI Overviews rank outside the top 50 organic results.Five-word queries trigger an AI Overview most frequently.Keywords from the Health and Safety niches are more likely to trigger AI Overviews.No AI Overviews show up for brand related queries.Navigational intent keywords are less likely to display AI Overviews.Google Ads are displayed in 28.3% keywords that trigger AI Overviews.From all the keywords that trigger AI Overviews, Ads at the top of the SERP appear for 8.7% of keywords. Ads at the bottom are displayed for 19.5% of these keywords.Shopping Ads are almost never seen together with AI Overviews and when they are, they always appear below the AI Overview.

AI Overview Visibility by IndustryWhich industries show AI Overviews more frequently?The bar chart below shows the frequency of AI overviews across different industries.

How to influence the organic product carouselIn Google’s blog post, they detailed three factors that are key inputs: Structured Data on your website, providing real-time product information via Merchant Center, along with providing additional information through Manufacturer Center.This section of the article will explore Google’s guidance, along with some commentary of what I’ve noticed based on my own experiences.

Make sure your product markup is validatedThe key here is to make sure Product Markup with Structured Data on your page adheres to Google’s guidelines and is validated.

Submit your product feed to Google via Merchant CenterThis is where it starts to get interesting. By using Google’s Merchant Center, U.S. product feeds are now given the option to submit data via a new destination.The difference here for Google is that retailers are able to provide more up-to-date information about their products, rather than waiting for Google to crawl your site (what happens in step 1).Checking the box for “Surfaces across Google” gives you the ability to grant access to your websites product feed, allowing your products to be eligible in areas such as Search and Google Images.For the purpose of this study we are most interested in Search, with the Organic Product Carousel in mind. “Relevance” of information is the deciding factor of this feature.Google states that in order for this feature of Search to operate, you are not required to have a Google Ads campaign. Just create an account, then upload a product data feed.Commentary by PPC Expert Kirk Williams:“Setting up a feed in Google Merchant Center has become even more simple over time since Google wants to guarantee that they have the right access, and that retailers can get products into ads! You do need to make sure you add all the business information and shipping/tax info at the account level, and then you can set up a feed fairly easily with your dev team, a third party provider like Feedonomics, or with Google Sheets. As I note in my “Beginner’s Guide to Shopping Ads”, be aware that the feed can take up to 72 hours to process, and even longer to begin showing in SERPs. Patience is the key here if just creating a new Merchant Center… and make sure to stay up on those disapprovals as Google prefers a clean GMC account and will apply more aggressive product disapproval filters to accounts with more disapprovals. ”– Kirk WilliamsFor a client I’m working with, completing this step resulted in several of their products being added to the top 10 of the PP carousel. 1 of which is in the top 5, being visible when the SERP first loads.This meant that, in this specific scenario, the product Structured Data that Google was regularly crawling and indexing in the US wasn’t enough on it’s own to be considered for the Organic Product Carousel.Note: the products that were added to the carousel were already considered “popular” but Google just hadn’t added them in. It is not guaranteed that your products will be added just because this step was completed. it really comes down to the prominence of your product and relevance to the query (same as any other page that ranks).

In this case, we concluded that there is a ranking boost, but the boost does not necessarily continue if you keep getting more reviews.

Small changes make a big impact. Google’s first point is that often with large sites, making small changes can make a big impact and return when it comes to search rankings. Google plotted the growth of one of the 7,000 websites, the Google My Business marketing site, showing how adding canonicals, hreflang to their XML sitemaps, and improving their metadata all resulted in gains in their organic traffic in search.Here is that chart:

Here is the chart showing the improvement after making the AMP error fixes:

Consolidation. For the past several years, many SEOs have been saying “less is more.” Meaning, having fewer sites and fewer pages with higher quality content often leads to better SEO results. Google says that works for them and they have been working on consolidating their sites. Google said they found a “large number” of near duplicate sites across their properties.“Duplicate content is not only confusing for users, it’s also confusing for search engines,” Google said. Google added, “Creating one great site instead of multiple microsites is the best way to encourage organic growth over time.”In one case study Google provided with the Google Retail site, they took six old websites and consolidated the content. They made “one great website” and it lead to them doubling the site’s call-to-action click-through rate and increased organic traffic by 64%.

Google has repeatedly said either is fine.

John Mueller said in 2017:Google websearch is fine with using either subdomains or subdirectories. Making changes to a site’s URL structure tends to take a bit of time to settle down in search so I recommend picking a setup that you can keep for longer.

Many SEOs believe that subdomains are treated as separate domains, but the truth is more complicated. Anyone that incorporates subdomains as a main part of their site will likely have them treated the exact same as a subfolder would be treated. However, if you’re not treating the subdomains as part of your main website (read as not connecting them with internal links), then they may be treated as separate.

To improve the topical Authority, a source should cover all the related details to a topic with a certain context, query, and intent template by satisfying the related and possible search intents. To improve the topical authority, the “keyword gap” is not as important as the “information gap”.

After cross referencing with the monthly search volume for these questions, she added some of these as H3s within the guide to see how they would perform. They resulted in so much more traffic that she decided they warranted their own individual guides, hence the idea for the hub. This would mean we weren’t putting all of our eggs into one basket, and that we could also internally link all of them together for users wanting to read more. Users that are further down the marketing funnel don’t want to scroll down a huge guide to find the answer to their specific question, and we were certain that this would positively affect bounce rate. We therefore made sure that nine times out of 10, the H1 contained the question that was being answered.

the content hub

LLM prompting involves designing text inputs to generate a response from an LLM. The goal of prompting is to steer the behaviour of the LLM in a way that elicits a desired outcome. Recent research has focused on developing effective prompting techniques that can expand LLMs' capabilities when carrying out a variety of tasks. Examples include prompt patterns [21], in-context instruction learning [22], evolutionary prompt engineering [23] and domain-specific keywords with a trainable gated prompt to guide toward a target domain for general-domain LLMs [24]. Zhong et al. [25] experiment with prompting LLMs to do scientific tasks across fields like business, science, and health by providing the LLM with a research goal and two large corpora, asking the LLM for corpus-level difference. Reppert et al. [26] develop iterated decomposition, a human-in-the-loop workflow for developing and refining compositional LLM programs that improves performance on real-world science question and answer tasks.