Does Schema Markup Affect Rank and AI Recommendations?

One of the challenges was finding people on both sides of the schema aisle to review the methodology of this study before it was conducted and the results were apparent. I wanted both doubters and hardcore believers to validate the methodology, suspending their bias.

Among the independent reviewers that answered the call to action, we had:

Jarno Van Driel

This is someone I’ve been dubbing “the grandfather of schema”. He worked directly with schema in 2008, which led him to work directly with Schema.org starting in 2011. He literally wrote the book on schema markup, specifically the schema.org markup we’re using in this study. He even works directly with the Yoast team on their schema implementation. Jarno not only reviewed the methodology you’re about to read in its entirety, but he also nitpicked the schema and ensured it was well formatted and that the most applicable types and IDs were used.

Jarno’s official stance?

Schema is valuable for contextualization and data aggregation, but does not influence search engine rank, nor LLMs.

Yoast (Alex Moss & Carolyn Shelby)

Speaking of Yoast, Edward Sturm was able to link me up with Principal SEO at Yoast, as Edward’s podcast was something we had both been on. Alex had also looped in Carolyn Shelby from the team, who is also Principal SEO over there.

Carolyn had provided the most extensive feedback on the methodology, including the reason the test was delayed by 3 weeks. We had originally planned to do a 2-week reset period in which both the Control and Test groups had no schema at all. However, she had suggested the reset period should be the same duration as the testing period.

Beyond reviewing the methodology, the role of Yoast in this study was important.

Alex and Carolyn reviewed the schema markup on each and every site in the Control and Test groups. Checking to make sure:

• Schema was accurate
• All sites had schema removed during the reset period
• Only the sites in the Test Group had the required schema during the testing period

Huge thank you to Alex, Carolyn, and Yoast on this one.

Alex and Carolyn’s official stance?

I’m honestly not sure. They seemed pretty independent; however, Yoast seems to take a stance in their documentation suggesting that it doesn’t help with search engine rank, only visibility, and that schema helps your site be understood by AI platforms.

Terry Samuels

Terry Samuels is the one staunch advocate of schema I got to review the methodology and meet with me over video call.

He’s been a technical SEO specialist for the last 30 years, running his own website and SEO agency, Salterra, for the last 15 years. Terry believes, through his own tests, that highly detailed structured data is necessary for search engines to know exactly what a business is and what they do, and why it holds authority.

Terry’s official stance?

“Advanced” schema is absolutely necessary in an SEO strategy.

David Quaid & Edward Sturm

I grouped these guys together because they’ve become a dynamic duo on Edward’s podcast.

Edward has become a notable name in the SEO community with his daily SEO show and short content. He’s consulted with Microsoft, P&G, Time, and more

David is a B2B SEO Growth Strategy Advisor specifically in SEO and PPC. He’s owned Primary Position since 2004. David’s an ultra-white hat SEO that puts his SEO theories to test and public critique (including my own). He spends his spare time managing one of the world’s largest SEO communities or guesting on other SEO podcasts.

David and Edward’s official stance?

Schema doesn’t impact rank. The most it does is qualify you for rich snippets and results. David also clarifies on this that some industries make schema a necessity, like travel and employment.

David Hunter

David Hunter is the founder and CEO of Epic Web Studios and LocalFalcon, the original local SEO geogrid tool, who graciously provided the credits required for this study. The entire reason this study happened is because of the discussion Cody and I had with him on our Agency Growth Podcast.

HUGE shoutout to David for helping with this study.

I also have to shout out his comment on the original methodology document, which gave me confidence to get more people involved.

David’s official stance?

David mentioned on his episode that he doesn’t think schema has anything to do with rank but thinks there is something to how schema is used in LLMs.

MOZ (Jonathan Berthold & Tom Capper)

LocalFalcon covered my local SEO need, but I needed SERP tracking help. That’s when, yet again, Edward Sturm connected me with Moz’ VP of Revenue, Jonathan Berthold, as both being alumni from Edward’s show. Jonathan then looped in Moz’s Director of Search Product Strategy, Tom Capper, to also review the methodology and give it the pushback it needed to be the best schema study out there.

I have to hand it to these guys and Moz. I don’t think they sleep. Every email got answered within a few minutes, no matter the time of day.

These guys hooked us up with a Moz Pro subscription and the white-glove treatment on setting up the SERP test within Moz.

Jonathan and Tom’s official stance?

Schema doesn’t impact rank, and they’re skeptical on LLMs.

Other Peer Reviewers and Honorable Mentions

Special shoutout to some other notable people who requested to review the methodology:

• Darren Shaw with Whitespark
• Joy Hawkins with Sterling Sky

Darren and Joy both share the same view. They’ve done their own tests on schema, and nothing has really indicated an increase in rank.

I would have loved to have Tim Soulo from Ahrefs review the methodology, but he’s been ignoring my emails since we had him on our podcast.

Also, Matt Diggity just outright refused to partake. Why, Matt? Don’t you love schema?

Pre-Test & Client Sorting

29 Clients (domains) with a total of 36 locations were selected. All within the same general industry (landscaping, lawn care, outdoor services). In our industry, we simply refer to this as the “Green Industry.”

Clients were sorted into two groups

• Control Group
• Test Group

All clients are located within the United States.

The groups were sorted based on geographic similarities to mitigate location and climate disparities and bias.

Control Group Domains (13)

• 6 South
• 3 Northeast
• 3 Midwest
• 1 West

Control Group GBP Locations (18)

• 6 South
• 3 Northeast
• 8 Midwest
• 1 West

Test Group Domains (16)

• 6 South
• 3 Northeast
• 6 Midwest
• 1 West

Test Group GBP Locations (18)

• 6 South
• 3 Northeast
• 8 Midwest
• 1 West

Additionally, clients who had started working with us in the last 6 months would not be included in this study. Initial efforts and site rebuilds that are relatively new cannot be considered due to potentially new rank delays from massive content and structural shifts.

Environment Cleaning & Benchmarking

All websites were scrubbed of any structured data and verified with Schema.org’s Schema Markup Validator.

All websites use Yoast SEO, which is the only element out of the plugins and themes used on our client sites that adds schema to the site. In version 26.8 of Yoast, we were able to toggle off the entire schema framework and validate that no schema was detected on sites whatsoever.

We had Yoast themselves validate this on all sites.

We then ensured all sites had a record of a Google Search Console crawl report after the schema was removed.

Control & Test Periods

A control period of 5 weeks was established to give time for sites and potential rankings to level out from any potential influences to prepare for this test. Simple things like the act of removing all schema.

Absolutely nothing related to the client’s SEO was touched during this period.

Following the Control period, a test period would commence, which would be the core of the test. It spanned 5 weeks and officially split the Control and Test Groups apart.

The Control Group continued to have nothing done to their SEO during the 5-week period.

The Test Group would have “advanced” LocalBusiness schema placed on their home pages.

Schema Placed

The schema below is an example of the same template we used for each client. All personal details have been removed.

Some details have been kept, such as hours of operation. This provides transparency on what formatting was used for things like OpeningHoursSpecification.

Additionally, some elements were consistent across all clients, such as the HomeAndConstruction type and the values for the NAICS and ISICv4 properties.

        <script type="application/ld+json">
        {
          "@context": "https://schema.org",
          "@graph": [
            {
              "@type": "HomeAndConstructionBusiness",
              "@id": "https://[clientdomain]/#organization",
              "name": "[client business name]",
              "legalName": "[client legal business name]",
              "mainEntityOfPage": {
                "@id": "https://[clientdomain]/#webpage"
              },
              "url": "https://[clientdomain]/",
              "logo": "https://[clientdomain/wp-content/uploads/client_logo.png]",
              "description": "Professional [description] services in the [target area].",
              "telephone": "+1-xxx-xxx-xxxx",
              "email": "[email]@[clientdomain]",
              "naics": "561730",
              "isicV4": "8130",

              "address": {
                "@type": "PostalAddress",
                "streetAddress": "[address",
                "addressLocality": "[city]",
                "addressRegion": "[state]",
                "postalCode": "[zip]",
                "addressCountry": "US"
              },

              "hasMap": "[google map link]",

              "geo": {
                "@type": "GeoCoordinates",
                "latitude": 	xx.xxxxx,
                "longitude": -xx.xxxxx
              },

              "areaServed": [
                {
                  "@type": "City",
                  "name": "[city]",
                  "containedInPlace": {
                    "@type": "AdministrativeArea",
                    "name": "[state abrreviation]"
                  }
                },
                {
                  "@type": "City",
                  "name": "[city #2]",
                  "containedInPlace": {
                    "@type": "AdministrativeArea",
                    "name": "[state abbreviation]"
                  }
                },
                {
                  "@type": "City",
                  "name": "[city #3]",
                  "containedInPlace": {
                    "@type": "AdministrativeArea",
                    "name": "[state abbreviation]"
                  }
                }
              ],

              "sameAs": [
  "https://www.facebook.com/[clientpage]",
                "https://share.google/[gbplink]",
                "https://www.yelp.com/biz/[yelpprofile]",
                "https://www.instagram.com/[igprofile]",
                "https://nextdoor.com/pages/[nextdoorprofile]",        
                "https://www.bbb.org/us/[bbbprofile]",        
                "https://www.mapquest.com/us/[mapquestprofile]",        
                "https://www.dnb.com/business-directory/company-profiles.[clientprofile]",        
                "https://[local city top 100 profile",        
                "https://www.glassdoor.co.nz/Location/[glassdoorprofile]",
                "https://yellow-pages.us.com/[yellowpagesprofile]"
              ],

              "openingHoursSpecification": [
                {
                  "@type": "OpeningHoursSpecification",
                  "dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday"],
                  "opens": "07:00",
                  "closes": "17:30"
                },
                {
                  "@type": "OpeningHoursSpecification",
                  "dayOfWeek": "Saturday",
                  "opens": "08:00",
                  "closes": "12:00"
                },
                {
                  "@type": "OpeningHoursSpecification",
                  "dayOfWeek": "Sunday",
                  "opens": "00:00",
                  "closes": "00:00"
                }
              ],

              "hasOfferCatalog": [
                {
                  "@type": "OfferCatalog",
                  "name": "Lawn Care Services",
                  "itemListElement": [
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Lawn Fertilization",
                        "description": "[brief description]"
                      }
                    },
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Weed Control",
                        "description": "[brief description]"
                      }
                    },
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Aeration",
                        "description": "[brief description]"
                      }
                    },
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Soil Tests & Amendments",
                        "description": "[brief description]"
                      }
                    },
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Insect & Pest Control",
                        "description": "[brief description]"
                      }
                    },
                    {
                      "@type": "Offer",
                      "itemOffered": {
                        "@type": "Service",
                        "name": "Fungus & Disease Control",
                        "description": "[brief description]"
                      }
                    }
                  ]
                }
              ]
            },
            {
              "@type": "WebPage",
              "@id": "https://[clientdomain]/#webpage",
              "url": "https://[clientdomain]/",
              "name": "[client name] | [location]",
              "isPartOf": {
                "@id": "https://[clientdomain]/#website"
              }
            },
            {
              "@type": "WebSite",
              "@id": "https://[clientdomain]/#website",
              "url": "https://[clientdomain]/",
              "name": "[clientname]",
              "inLanguage": "en-US",
              "publisher": {
                "@id": "[clientdomain]/#organization"
              }
            }
          ]
        }
        </script>

Considerations

The LocalBusiness schema above was only placed on the home page of a client’s website in accordance with Google’s general structured data guidelines.

“Schema Impacts Rank” Not “Which Schema Impacts Rank”

The aim of this outcome is to elicit a simple yes or no answer. To accomplish this, the test needed to focus on standard, non-rich-snippet-producing schema.

If schema did impact rank, then a separate test for “which schema impacts rank” would be considered.

Article, Author, and Other Schema Types Will Not Be Used

Not all clients have a robust blog or a blog at all, which includes authors or other relevant options to add this schema. Adding a blog and authors to each site would skew the results and result in inconclusiveness due to a lack of variable isolation with the added content.

Review Markup Will Not Be Used

Simply put, in most cases, adding review markup schema to a local business site goes against Google’s guidelines.

For the safety and future success of our clients, we will avoid this.

FAQ Schema Will Not Be Used

Google eliminated the ability for FAQs to show in rich results in 2023, except for government or health-related organizations.

Shortly after this test concluded, Google announced the deprecation of FAQ rich snippets altogether on May 7th, 2026.

Some will argue that doesn’t mean FAQ schema is no longer important. My rebuttal would be if Google no longer needs it to show rich results, then why is it needed at all?

Rich Snippets To Be Avoided

An argument can be made for schema that it can indirectly improve rank by improving click-through-rate via rich snippets. For the purpose of this test, our schema implementation is designed to avoid all rich snippets, including breadcrumbs.

Whether schema markup influences CTR and by how much is to be the subject of a different test.

Search Engines and LLMS Tested

There were two primary tools used for this study.

Moz Pro for SERP analysis on the following search engines:

• Google
• Google Mobile
• Bing
• Yahoo!

LocalFalcon for Google Maps and LLM analysis on the following:

• Google Maps
• Google AI Mode
• Google AI Overviews
• Gemini
• Grok
• ChatGPT

Queries & Metrics Tracked

The entirety of the test reviewed rank and placement of two queries:

Query #1
“[service] company in [target location]”

Query #2
“can you recommend a [service] company that services the [location] area that’s open at noon on friday?”

Query #1 was the primary target for SERP analysis, but we wanted to see if search engines behaved differently when longer queries were used that called for objective information. Objective information that can be found in schema.

Query #2 was also a topic of interest for the LLM testing side of things with LocalFalcon. It was also included in the SERP analysis for research consistency.

After the 5-week control period, during which schema was stripped from all sites and SEO efforts were halted, a benchmark report was conducted on the final week of the 5-week control period on all platforms and tools. After which, the LocalBusiness schema was added to all Test Group sites, and Google Search Console crawl reports were validated after the fact to ensure the home pages had been recrawled and indexed.

The results would be captured and plotted for each client, for each keyword, and each platform week-over-week.

*The week-over-week data capture was merely a failsafe to ensure there weren’t any reporting issues throughout the 5-week testing period. Because we had this failsafe in place, it actually caused us to switch benchmarking for LocalFalcon to Friday, February 27th, instead of the originally planned Wednesday, the 25th. For some reason, Grok was not recording results on Wednesdays but would output answers sufficient for data collection on Fridays.

SERP Testing with Moz Pro

SERP results were captured using Moz Pro. One campaign was created for each client in the test, which tracked the average position of both of our target keywords week over week across the four search engines.

Both average rank and Search Visibility would be plotted each week until the end of the 5-week testing period.

As a refresher:

• Rank
  • The position our client is in for their target keyword in the organic results (aside from the Mapp pack).
• Search Visibility
  • The total estimated percentage of clicks the client is likely to get based on their organic ranking positions for all tracked keywords.

No other SEO changes would take place on any domain or GBP.

Google Maps & LLM Testing with LocalFalcon

To determine local rank and local recommendations by LLMs, we used a 9×9 grid and tested each query at each plot point weekly.

That’s:

• 2 Queries
• x81 Plot points
• x36 GBP locations
• x6 Weeks (includes the benchmark week)
• x6 Environments

Each query for each client on each platform received 486 searches throughout the 6-week period.

Due to LocalFalcon’s advanced brand and entity recognition, an LLM “rank” metric can be used in this test.

Similar to the testing methodology with Moz, we set up 36 individual campaigns and monitored them week over week, capturing the final results on April 3rd, 2026.

During this 5-week testing period (plus the benchmark week), we reviewed the average results from three primary metrics on all platforms:

• Average Total Rank Position (ATRP)
  • Measures the average total rank of all 81 query searches in the gridded target area.
• Share of Local / AI Voice (SoLV / SAIV)
  • SoLV measures the overall percentage your brand appears in the top three Google Maps results for the query you’re targeting. SAIV essentially does the same thing for AI / LLM recommendations. The only difference is it doesn’t count whether you’re in the top three or not. It is whether your brand is included in the recommendation or not, and the percentage at which it does.

Analyzing the Results

Most studies will simply suggest that if the average rank of the Test Group improved over the average rank of the Control Group, then the variable in the Test Group was the root cause. However, that is an oversimplification and doesn’t account for uncontrollable factors between the groups.

In order to be confident in the analysis between the groups, a One-Tailed Welch’s Two-Sample T-Test was conducted between the deltas of each and every client’s benchmark result against their final result to determine a degree of confidence.

If the Test Group outperformed the Control Group by 3 positions, the T-test will indicate how mathematically confident we are that the schema was the result of the change in the Test Group over the Control Group.

One thing we have to consider is that the one-tailed T-test only detects confidence in change and not improvement. For this test, we only want to know if an improvement happened.

This means that the average increase of the rank over the 5-week period must be greater in the Test Group than the Control Group in order to consider running the test.

If the Test Group did not outperform the Control Group, then we have to assume that the variable (schema) had no impact, and we place a “0%” in the confidence field.

If the Test Group does outperform the control group and a T-test is run, we must accept a 90% or higher confidence percentage to make definitive claims that the schema was the cause of the improvement.

For purposes of understanding the confidence level, 50% is akin to a coin flip and is not statistically significant. The same can be true for anything between 25% and 75%. As a marketer considering realtively low effort tasks with no real downsides, I’ll generally accept something at 80% for my own efforts but will not promote anything under 90% as fact.

My rule of thumb:

• <50% = “No.”
• 50%><80% = “Probably not.”
• 80%><90% = “Maybe. Why not? But focus on other things first.”
• >90% = “Yes. You should do this.”

One-Tailed Welch’s Two-Sampled T-Test

The formula for this is expressed like this:

In our Google Sheet, this is what the formula looks like:

The numbers are pulled from the deltas of the benchmark and final results of each individual client in the Control Group and are run against the Test Group deltas. This is how the confidence levels are determined.

Keep in mind that a 0% means the average of the test group did not perform better than the average of the Control Group during the testing period, and a null 0% was entered.

Potential Criticisms

Any information presented to people that is contrary to their existing beliefs and expectations is met with trepidation and is subject to additional scrutiny.

This section aims to address those based on my own observations.

Only Testing Local Service Area Businesses in One Industry

Testing by isolating one specific industry is the best way to test. That is, removing the industry variable as schema could also affect industries differently. If there was an industry mix in the control and test groups, that could skew the data.

I welcome people to “steal” this methodology and test it on other industries.

This conclusion to this test will stress the importance or lack thereof of structured data markup for a particular industry.

Not Using All Possible Structured Data

This too could skew the results, as some clients and businesses do not have some features others have. The structured data chosen was data relevant to all businesses.

Additionally, the aim of the test is to extract a simple, boolean answer, not an array of which schema affects rank.

Algorithms Change Over Time and Will Influence Results

This is why the groups are split into a control group and test group.

The groups will move in parallel through time and algorithm updates. If both groups increase rank, the question becomes, “Did one group increase more in rank than the other?” The opposite direction would be true as well.

Seasonality of Industry

The landscaping industry sees a pretty heavy seasonal change heading into Q2. CTRs, impressions, and even rank can be influenced strictly by an increase in localized demand.

Again, the test and control groups solve for this.

In addition, structuring each group so they share geographic similarities reduces the variance in weather and climate influences on the total sample sizes.

Uncontrollable Factors

Although we manage the website and SEO for all clients in this study and are consciously halting all SEO efforts, there are still uncontrollable factors.

These factors include things like:

• Review quantity, frequency, and ratings
• Social media influences
• Local influences
• Competitor influences
• etc

This is the reason the Welch test is being used. It was specifically designed for real-world variance between two groups.

My Hypothesis Presents a Bias

Many people know my views on schema, and some may believe that this test or the results can be manipulated to seek the answer I desire.

However, I have four points to this:

1. This methodology is public.
2. This methodology has been peer-reviewed by members of varying beliefs and positions related to schema prior to the release of the results.
3. The schema used and placed on the sites (or lack thereof in the Control Group) was reviewed and validated by Alex Moss and Carolyn Shelby of Yoast throughout the test.
4. I would love for the antithesis to be proven.

The final point is the most important.

Our agency does not currently implement schema on our client websites. If schema does, in fact, show promise or significant influence on search and LLMs, then this adds value to our service by way of including this line item.

… and the study to prove it.