EspañolSEO Research for Decisions, Not Assumptions
SEO research can look convincing while hiding the baseline, the measured pages or the type of traffic counted. This library puts methods beside the headline so you can judge what a result means for people and for your business.
SparkSEO, a performance-oriented SEO agency in Barcelona, shares its published results alongside independent studies and clear caveats. Use these findings with our wider SEO resources to set better questions, rather than treating a single chart as a promise.
- Published client data
- Clear measurement rules
- Independent study summaries
- AI search context
- Practical reading checks
Based on 14 active clients with at least 12 months of data. Updated October 2026. See how we measure results
The published SparkSEO SEO statistics, in context
A headline result can invite the wrong conclusion when it does not state who was measured. For example, an average can be pulled upward by one exceptional account, so SparkSEO reports a median result to describe the middle outcome rather than a claim about every client.
The homepage summary is based on 14 active clients with at least 12 months of data and was updated in October 2026. These are published operational results, not a controlled experiment or a forecast, and the fuller context belongs with results and case studies.
How the 12-month result summary is measured
SEO claims can blend brand demand, paid campaigns and many pages into one number. Imagine a company with a well-known name: more people searching for that name can increase clicks even if its service pages did not become easier to find, so that growth does not answer a non-branded SEO question.
SparkSEO therefore measures one approved group of related keywords for each managed landing page. This page-level approach connects research to the practical work in keyword and AI prompt research, where intent and relevance matter more than a large, disconnected keyword list.
The published click figure excludes branded searches, meaning searches that include the client company name. Non-branded clicks are useful because they better show whether people who do not already know the company can discover a relevant page. They still do not prove that SEO alone caused every change, because seasonality, demand and search-result changes can affect visits.
Each landing page is evaluated as its own unit. SparkSEO works with one landing page and one keyword group, then records position change monthly. This makes a weak page visible instead of allowing a strong page to conceal it, and it helps a team decide whether content, technical work, internal links or authority work is the next reasonable priority.
The 71% figure is a median. Place the measured landing-page outcomes in order, then take the middle value. A median limits the influence of unusual highs or lows, but it does not show the full spread of outcomes, so it should sit beside page-level reports and the number of pages measured.
For a decision, ask to see the definitions, the reporting period and the work recorded alongside the movement. SparkSEO provides weekly and monthly updates, and its SEO reporting approach separates performance measures from progress work. Full anonymized reports are available on request.
Anonymized cases show the work behind the numbers
An aggregate statistic can conceal very different problems. A dental group may need accurate local details, while a multilingual store may lose visibility through incorrect language targeting, so a useful SEO study should name the context and the work performed.
The following published cases are anonymized because client names are private under NDA. They are examples of reported outcomes, not predictions for another business, and their different starting points are the reason a single tactic cannot fit every site.
Dental clinic group, Barcelona
Three locations were measured over 12 months for local SEO. Calls from Google Business Profile rose 61%, and 19 of 48 keywords reached the Map Pack Top 3, up from six. The work covered Google Business Profile, location pages and reviews. For a sector example, read the SEO for dentists guide.
Furniture e-commerce, seven markets
A five-language store recorded 48% growth in non-branded organic revenue over 12 months. Hreflang errors fell from 1,180 to 36 after technical SEO, hreflang and content work. A May decline after a Google core update was followed by recovery, which shows why a time series matters.
B2B HR software, US and UK
Over 10 months, client citations appeared for 11 of 35 tracked AI prompts, up from two, while demo requests from organic search rose 34%. The work combined an AI prompts audit, content and blogger outreach. The result concerns this prompt set and market, rather than every AI answer.
Pew data shows why AI visibility needs a traffic check
An AI answer can make a brand visible without sending a visit, which complicates a simple ranking or mention report. For example, a page may be cited in a summary while the user ends the session, so visibility, clicks and conversion need separate measurement.
In Pew Research Center's 2025 analysis, browsing data from 900 U.S. adults covered 68,879 Google searches in March 2025. The study found AI summaries in 18% of those searches, which supports tracking referral traffic and meaningful actions alongside citations.
Users who encountered an AI summary clicked on a traditional search result link in 8% of all visits. Those who did not encounter an AI summary clicked on a search result nearly twice as often (15% of visits).
The GEO paper offers a hypothesis, not a shortcut
AI search claims often turn a research finding into a promised outcome. For example, a visibility lift in a benchmark can depend on the domain, the engine and the test conditions, so it cannot stand in for a client forecast.
The GEO paper accepted to KDD 2024 reports that its methods could boost visibility by up to 40% in generative-engine responses and that effectiveness varied across domains. This AI search study is useful for forming questions about citations and presentation, but its own caveat calls for domain-specific testing.
The practical response is to start with content that earns trust from a person. State the question, show the evidence, explain who is responsible and keep key business details accurate. In Google's people-first guidance, Google asks creators to consider who made content, how it was made and why it exists, which is a sound test for a research page as well as an AI answer.
Technical clarity still matters. Google's AI features guidance says its generative AI features use core Search ranking and quality systems. It also says a special AI text file or special schema is not required, so mass-producing query variants is a poor substitute for relevant expertise and clear pages.
For a company, AI research should therefore define the prompts, country, language, date and engines checked before a result is discussed. The next step is to compare citations with qualified visits and outcomes, then improve the source page where the user experience has a genuine gap. Our AI Search Optimization service applies that sequence without treating mentions as an end in themselves.
How to read SEO statistics critically
SEO statistics can sound precise while answering a different question from yours. A graph might show more impressions after a site relaunch, for example, but omit whether non-branded traffic, leads or revenue changed, so the reader needs a small set of checks before acting on it.
The solution is to examine scope, comparison and outcome in that order. These checks make it easier to value transparent work and to avoid shortcuts that add activity without relevance for a human visitor.
- Define the population Look for the number and type of sites, pages or queries measured. A cohort of active clients is informative only when the reader knows the inclusion rule, time requirement and markets represented.
- Check the measurement unit Ask whether the result is per domain, landing page, keyword, search visit or conversion. A page-level measure is more actionable because it can point to a specific content or technical issue.
- Find the baseline and period A percentage needs a clear starting date, end date and baseline. Compare enough months to see seasonality or a documented search update rather than reading one short movement as a trend.
- Identify the summary statistic A median describes the middle observation; an average can be moved by extreme values. Ask for the range or page-level distribution when a summary is used to make a budget decision.
- Separate association from cause Traffic can move because of demand, competition, site changes or a search-system change. A responsible report records the work done and the timing, then avoids saying that one action caused every result.
- Connect to a human outcome Higher positions matter when the landing page helps a relevant person call, enquire, buy or learn. For terminology that keeps reporting conversations clear, use the SEO glossary.
Research methods to define before reporting new findings
A research page can become less trustworthy when it presents an intention as a completed study. For example, an AI prompt review has no result until its prompts, locations, models and checking method are defined, so planned work needs a visible label.
SparkSEO will publish the method, sample and limitations before reporting new findings. The three guides below explain the questions a responsible study must answer; they are research-method resources, not announcements of active studies or predicted results.
- Research method
AI prompt visibility audit
Before reporting AI visibility, define the prompts, country, language, AI systems, checking dates and citation rule. Then compare citations with qualified visits and useful actions so a mention is not mistaken for traffic or a lead.
- Research method
Landing-page performance cohort
Before comparing landing-page movement, define the page types, approved non-branded keyword groups, baseline and measurement period. Record the work completed alongside changes in clicks, positions or conversions so the results can be interpreted rather than assumed to have a single cause.
- Research method
Reporting transparency review
A useful reporting review should specify the measures shown, the review method, the reporting period and which evidence can be shared with permission. This makes it easier to separate recorded progress from performance outcomes and to understand the limits of the comparison.
Terms that make an SEO study easier to assess
Research language can hide a practical choice. A team may approve a report without knowing whether a click was branded or whether the central figure was a median, which makes later comparisons unreliable.
These short definitions establish a shared vocabulary before a result is used in planning. The wider SEO guides explain the related work in more detail.
- SEO research
- The organised collection and interpretation of search, page and user data to decide what will help a relevant person find and use a site.
- Non-branded click
- A click from a search that does not include the company or product name. It is often used to assess discovery beyond existing brand recognition.
- Landing page
- The page a searcher reaches from a result. In a sound study, its purpose and keyword group should be clear.
- Median
- The middle value after observations are put in order. It is less affected by an unusual high or low result than an average.
- AI mention
- A client name or source appearing in a tracked AI response. It indicates visibility for that prompt set, not a guaranteed visit, lead or sale.
- Generative engine optimization
- A research term for improving visibility in generative-engine responses. The GEO paper presents it as a benchmarked optimisation framework, while real-world outcomes need independent checking.
Questions about SEO research
What does SEO research include?
SEO research combines search demand, user intent, page performance, technical access and competitor context. A useful programme turns those findings into page-level priorities, then records the work and outcome so the next decision has evidence behind it.
Why does SparkSEO use non-branded clicks?
A non-branded search does not include the client name, so it helps show discovery by people who may not know the company already. Brand growth can still be valuable, but it answers a different question and should be reported separately.
Does a 71% median result mean every client grew by 71%?
No. A median is the middle landing-page outcome in the published measurement set, not an outcome promised to every client. The sample, page types, baseline and conditions should all be reviewed before comparing it with your site.
What should an AI search study measure?
It should define the prompts, country, language, AI system, checking date and what counts as a citation or mention. It should also compare visibility with qualified visits and useful actions, because an AI mention does not automatically create traffic.
Are AI search mentions a replacement for SEO reporting?
No. Google's generative AI features are rooted in its core Search ranking and quality systems, according to its official AI optimization guide. A complete report still needs pages, queries, clicks, technical health and the human outcome that matters to the business.
How can I compare SEO studies from different providers?
Compare the population, measurement unit, baseline, period, statistic and stated limitations before comparing the result. If the method is absent, the figure may still be interesting, but it is too weak to guide a confident decision.
Sources
- Pew Research Center: Google users are less likely to click on links when an AI summary appears in the results
- Aggarwal et al.: GEO: Generative Engine Optimization, accepted to KDD 2024
- Google Search Central: Optimizing your website for generative AI features on Google Search
- Google Search Central: Creating helpful, reliable, people-first content
Get Your Free SEO Plan and Price
Share your website and your personal SEO manager will review it. You receive a plan with priorities, the keyword groups we would target first and a price that follows your non-branded rankings.
- Monthly billing. Cancel at any time and keep every result you paid for.
- A personal SEO manager who is online during Barcelona working hours.
- Weekly and monthly reports measured per landing page on non-branded keywords.