A year of SearchPilot split tests on adding FAQ schema

Most of the industry commentary on FAQ schema treats it as a single lever: add it, get a bigger snippet, get more clicks. SearchPilot’s own testing across a year of client work complicates that story in a useful way. The markup won more often than it lost, but the size of the win moved with the device, the wording of the questions, and what other markup was already competing for the same SERP space.

What the tests actually measured#

SearchPilot’s July 2020 case study updates an earlier report from August 2019, folding in a wider run of controlled split tests on FAQ schema “across various clients across various industries.” Each test follows the same structure the company uses across its case studies: similar pages on a live site are split into a control group, which keeps its existing markup unchanged, and a variant group, which gets FAQPage schema added to FAQ content the page already displayed. Both groups run at the same time, on the same site, so the only difference between them is the markup itself, not the calendar period. SearchPilot compares the variant group’s actual organic traffic against a statistical forecast of what it would have done without the change, and reports the result with a confidence interval rather than a raw before-and-after percentage.

What it found#

Sixty-seven percent of the FAQ schema tests SearchPilot has run came back positive, which the company describes as its most consistently positive test type across everything it publishes. Positive results typically landed between a 4% and 15% uplift in organic traffic, wider than the 3% to 8% range SearchPilot had reported a year earlier, with one client test reaching a 25% uplift from adding FAQ schema to FAQ content that was already live on the page.

The one negative result came from a markup trade-off, not the schema itself. SearchPilot found that Google would award only one of FAQ or review/price rich results on a given page, not both. One test removed review and price schema specifically to let FAQ schema win the SERP feature instead, and that swap produced a 5% decline in organic traffic. SearchPilot reads this as evidence that review and price data carries more weight than FAQ data for ecommerce-style pages where review content is relevant to the user, rather than as a mark against FAQ schema on its own.

Device and question wording both moved the number. Uplifts ran consistently higher on mobile than desktop across the positive tests, which SearchPilot attributes to the FAQ rich result claiming proportionally more of a smaller mobile screen. On one client, rewriting the first FAQ question to target a higher search-volume keyword, while keeping the schema itself unchanged, produced a further statistically significant 2% uplift in traffic and a 0.2 percentage point lift in conversion rate, a reminder that the questions themselves still function as content, not just markup.

Why this is a 2020 result, not a current one#

Nothing about this case study’s mechanism survives to today. The uplift in every positive test came from winning the FAQ rich result and the extra SERP space it occupied, and Google no longer grants that result to anyone. Google restricted FAQ rich results to authoritative government and health sites in August 2023, then removed the feature for those remaining sites too on 7 May 2026, as covered separately. A split test run today cannot reproduce these percentages through the route that produced them, because the SERP feature they were competing for is gone.

What survives is the shape of the evidence, not the numbers. SearchPilot’s separate, later test on removing FAQ schema after the rich result disappeared found no traffic effect either way, which is consistent with this earlier data: the traffic gain here tracked winning a visible rich result, and once that result stopped existing, so did the traffic reason to keep the markup. The two case studies, read together, describe the same feature’s full arc, a real uplift while the SERP feature existed, and nothing measurable once it did not.

Why we care#

Schema case studies age unevenly. A controlled split test is stronger evidence than an observational before-and-after comparison at the moment it runs, but the result it produces is tied to whatever SERP feature the schema was competing for. This one is worth keeping on record precisely because it shows what FAQ schema did do, with real controlled numbers, before the feature it depended on was taken away. Anyone weighing whether a still-active piece of structured data is worth the implementation effort should read case studies like this one for the method, not the percentages, because the percentages have an expiration date the method does not.