Research

Research data catalogue

Every file I publish, with what it holds, the window and sample behind it, what I left out, how I checked it and how to cite it.

Catalogue entries
21
Downloadable files
26
Last checked

What licence applies?

I license every file on this page under CC BY 4.0, except the script (MIT) and the third-party figures in compiled sets, which belong to their sources. Reuse the data with credit to Sunny Patel and keep the window and sample with each figure.

On compiled files the licence covers my selection and wording. The figures stay with the source named on each row.

Do the files name any sites?

No file names a site that I own, apart from this one. On 2 November 2026 I checked the public files for site names.

The link building file names sunnypatel.co.uk. The AI Overviews JSON names third-party sites that Google cited. Both are marked on their entries. One table inside a study, not a file, names my own sites. It is marked too.

Data I collected

Files from my own pulls, with a download for each.

  • AI referral traffic, edition 2

    What it contains
    One row per GA4 property: sector, anonymous site number, total sessions, organic sessions, sessions from ChatGPT, Copilot, Perplexity, Claude and Gemini, AI total and AI share. The JSON adds sector splits, panel totals and the sensitivity check.
    Window
    24 June to 21 September 2026, in each property's own GA4 time zone
    Study page
    AI referral traffic studyDated
    Sample
    108 owned GA4 properties. 104 recorded sessions and form the analysis panel.
    Exclusions
    Client sites, 2 accounts I do not publish from, and 7 further properties removed in a 25 September ownership review (2 clients, 5 with ownership I could not confirm), chosen on ownership and not on their numbers. One duplicate property. Four properties with zero sessions stay in the file but not in the shares.
    Anonymisation rule
    Sector label and an anonymous site number such as S001. No site names.
    Site names in the file
    None found.
    How it was verified
    One script applied the matching rule to one stored raw pull per property. A second, public Python script recomputes the headline table from the JSON offline and prints the file's SHA-256 hash.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). AI referral traffic, edition 2 [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/ai-referral-traffic-study/ Licence CC BY 4.0.

  • AI referral traffic, edition 1

    What it contains
    AI sessions by sector for 12 sites: total AI sessions and sessions from ChatGPT, Copilot, Perplexity, Claude and Gemini.
    Window
    28 May to 26 August 2026 (91 days)
    Study page
    AI referral traffic studyDated
    Sample
    12 sites I already knew received the most AI traffic. This is a hand-picked panel and not a sample of the portfolio.
    Exclusions
    Every site outside the 12. No denominators are in this file, which is why edition 2 exists.
    Anonymisation rule
    Sector label only. No site names.
    Site names in the file
    None found.
    How it was verified
    No script. Re-pulling the original window later gave 4,745 AI-matched sessions across the 12 properties, against 4,717 published. The mapping from old labels to current properties carries some uncertainty.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). AI referral traffic, edition 1 [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/ai-referral-traffic-study/ Licence CC BY 4.0.

  • UK AI Overviews by industry

    What it contains
    Per sector: panel size, usable queries, queries with an AI Overview, presence rate, citation counts by source type and the largest share supplied by one seed topic. The JSON adds overall totals, citation categories, the most-cited domains and seed concentration.
    Window
    One extraction on 25 September 2026
    Study page
    UK AI Overviews by industryDated
    Sample
    A locked panel of 399 UK queries across 10 sectors, built from 5 seed topics per sector. 393 returned usable data. A fixed random seed drew a 100-query citation sample, 10 per sector.
    Exclusions
    6 queries with an empty feature result, treated as missing. 17 redirect-labelled citation rows, left out of resolved totals. Google Shopping rows. Queries under 100 UK searches a month.
    Anonymisation rule
    The queries are not personal data. The JSON names the public websites that AI Overviews cited.
    Site names in the file
    The JSON names third-party sites that AI Overviews cited, such as gov.uk. None of them is one of my sites.
    How it was verified
    The panel and the random seed were fixed before any data was pulled. I did not write a second script for this study.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). UK AI Overviews by industry [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/uk-ai-overviews-by-industry/ Licence CC BY 4.0.

  • UK SEO consultant fees

    What it contains
    Aggregate prices by bucket: number of providers, median, quartiles, minimum and maximum for monthly retainers, audits, local SEO and link building, nationally and by region group. The JSON also lists the raw low and high price points.
    Window
    24 to 25 September 2026
    Study page
    SEO consultant fees in the UKDated
    Sample
    75 UK providers that publish their own price, from 150 UK provider sites screened. 72 are agencies and 3 are freelancers.
    Exclusions
    12 non-UK domains, 17 pages that returned an access error, 75 UK sites with no qualifying own price, directories and roundups. My own pricing page is not in the panel. A bucket needs 8 providers before it gets a median.
    Anonymisation rule
    Aggregates and price points only. Provider names and URLs are not published.
    Site names in the file
    None found.
    How it was verified
    The inclusion rules and the 8-provider minimum were set before the figures were calculated. The fetch ledger and compute script are kept on file. I did not write a second script for this study.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). UK SEO consultant fees [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/how-much-does-an-seo-consultant-charge-uk/ Licence CC BY 4.0.

  • Google CTR by position, July 2026 collection (withdrawn)

    Withdrawn
    What it contains
    CTR, clicks, impressions and query-row counts by position bucket, with the collection rules and limits repeated on every row.
    Window
    9 April to 7 July 2026
    Study page
    Google CTR by positionDated
    Sample
    53 qualifying Search Console properties in a convenience portfolio. 2,615 query rows.
    Exclusions
    Queries under 50 impressions. Only the top 5,000 rows per property, ordered by clicks. There was no country filter and no brand exclusion.
    Anonymisation rule
    Bucket totals only. No per-property rows and no site names.
    Site names in the file
    None found.
    How it was verified
    Withdrawn on 25 September 2026. The position-one bucket held 5 query rows. A re-extraction with a UK filter and brand exclusion also failed to support a benchmark. The file stays online for the audit trail. Do not cite it as a benchmark.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). Google CTR by position, July 2026 collection (withdrawn) [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/google-ctr-study/#correction Licence CC BY 4.0.

Data I compiled from other sources

Published statistics I checked and put in one file. The figures belong to their sources.

  • UK SEO statistics 2026

    What it contains
    18 sourced statistics: value, statement, period, source date, sample, source, source link and caveat. The JSON adds the zero-click chart data.
    Window
    Source dates run from 22 July 2025 to 23 October 2026. Every row was re-checked on 23 October 2026.
    Study page
    UK SEO statistics 2026Dated
    Sample
    Figures published by other organisations, such as the ONS and SparkToro. The sample behind each figure is stated on its row.
    Exclusions
    Any figure I could not trace to a source I opened.
    Anonymisation rule
    Not applicable. The rows are published statistics, with no site or client data.
    Site names in the file
    None of my sites. The rows link to the publishers.
    How it was verified
    Each row was re-checked against its source on the date in the checked column.
    Licence
    CC BY 4.0 for my selection and wording. Each figure belongs to the source named on its row. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). UK SEO statistics 2026 [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/seo-statistics-uk/ Licence CC BY 4.0.

  • UK marketing salaries 2026

    What it contains
    ONS median and mean annual gross pay for marketing occupations, with year-on-year change and the ONS quality key. National and regional tables.
    Window
    ONS ASHE 2026 provisional, pay period that included 22 April 2026. The regional table is ASHE 2025 provisional.
    Study page
    UK marketing salary statisticsDated
    Sample
    Full-time employees, from the ONS Annual Survey of Hours and Earnings. The sample belongs to the ONS.
    Exclusions
    Part-time employees. Northern Ireland in the regional table, which the ONS table does not cover.
    Anonymisation rule
    Not applicable. The file holds ONS aggregates.
    Site names in the file
    None.
    How it was verified
    Checked against the ONS tables on 23 October 2026. The 2026 estimates are provisional.
    Licence
    CC BY 4.0 for my selection and wording. Each figure belongs to the source named on its row. The ONS figures are published under the Open Government Licence v3.0. Licence text Open Government Licence v3.0
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). UK marketing salaries 2026 [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/uk-marketing-salary-statistics/ Licence CC BY 4.0.

Tables inside studies, with no file yet

These tables live on the study page. Cite the page until a file exists.

  • Google update tracker table

    File: none yet. The tables are on the study page.

    What it contains
    For each confirmed core and spam update: eligible sites, the number up or down by 20% and 50%, the median change and the pooled change.
    Window
    June 2025 to October 2026
    Study page
    How my sites changed after each Google updateDated
    Sample
    100 owned Search Console properties. Eligible sites per update run from 0 to 66.
    Exclusions
    Sites without data before the before window, or with under 100 impressions in it. Client sites. Median and pooled change are hidden when fewer than 30 sites are eligible.
    Anonymisation rule
    Counts and percentages only. No site names.
    Site names in the file
    None in the table.
    How it was verified
    A second script, written separately, recomputes every row. All nine rows match on n, counts, median, pooled change and both sums.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). Google update tracker table [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/google-update-portfolio-impact/#update-tracker Licence CC BY 4.0.

  • Query length tables

    File: none yet. The tables are on the study page.

    What it contains
    Share of impressions, clicks and query rows by word count, CTR by word count, top-3 zero-click share and question-query share.
    Window
    30 October 2025 to 29 October 2026
    Study page
    Search query length on 97 small sitesDated
    Sample
    97 of 100 owned sites. 223,218 query rows, 2,717,358 impressions and 12,551 clicks.
    Exclusions
    Queries Google withholds, which are rare and skew long. 3 sites returned no query rows. Client sites.
    Anonymisation rule
    Pooled counts and shares. No queries and no site names are published.
    Site names in the file
    None in the tables.
    How it was verified
    A second script, written independently, matched every headline number exactly.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). Query length tables [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/search-query-length-study/ Licence CC BY 4.0.

  • Small-site benchmark tables

    File: none yet. The tables are on the study page.

    What it contains
    Pages with impressions per site, zero-click page share, click concentration, new-page timings to first click and size-band medians.
    Window
    30 October 2025 to 29 October 2026
    Study page
    Small-site SEO benchmarksDated
    Sample
    100 owned sites out of 183 property rows. Site medians use 84 sites.
    Exclusions
    Client, lead and family sites, properties with unclear ownership, duplicate properties and sunnypatel.co.uk, whose migration distorts the new-page measure.
    Anonymisation rule
    Medians, quartiles and pooled counts. No site names.
    Site names in the file
    None in the tables.
    How it was verified
    A second script, using separate pulls, matched all 5 headline numbers exactly. That checks the arithmetic and not Google's own sampling.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). Small-site benchmark tables [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/small-site-seo-benchmarks/ Licence CC BY 4.0.

  • AI traffic across 20 sites

    File: none yet. The tables are on the study page.

    What it contains
    AI-referral sessions and share of sessions for a sample of owned sites, by assistant.
    Window
    21 June to 18 September 2026
    Study page
    How much traffic comes from AI search: 20 sitesDated
    Sample
    20 owned GA4 sites out of around 115 properties, the figure the study page gives.
    Exclusions
    The other properties. One site is shown with and without its anomalous Direct traffic.
    Anonymisation rule
    The study table names owned sites. It holds no client data.
    Site names in the file
    Yes. The study table names my own sites.
    How it was verified
    I did not recompute this table independently. The AI referral traffic study supersedes it for portfolio-wide shares.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). AI traffic across 20 sites [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/ai-search-traffic-portfolio-data/ Licence CC BY 4.0.

  • Multimodal filter table

    File: none yet. The tables are on the study page.

    What it contains
    Multimodal impressions, share of web impressions and CTR for three anonymised retailers.
    Window
    10 to 25 September 2026
    Study page
    Search Console multimodal filterDated
    Sample
    Three client properties with meaningful volume, out of 46 accessible properties.
    Exclusions
    Properties with no multimodal impressions are counted in the text, not the table.
    Anonymisation rule
    Retailers are described by type. No names.
    Site names in the file
    None in the table.
    How it was verified
    For one retailer, the interface view and the API returned the same Web rows, which confirms the API's Web figures exclude multimodal.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). Multimodal filter table [Data set]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/search-console-multimodal-filter/ Licence CC BY 4.0.

Templates, briefs, scripts and reference lists

Everything else in the downloads folder. None of it is research data.

  • AI referral sources for GA4 (reference list, version 1.0)

    What it contains
    The 15 referrer sources I count as AI assistant traffic in GA4, the sources I watch without counting, and the date each was added. It is a reference list, not a study.
    Window
    Not applicable. Version 1.0, 2 November 2026.
    Study page
    AI referral sources in GA4Dated
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    No data.
    Site names in the file
    None found.
    How it was verified
    Not applicable. The study page compares the list's total with GA4's AI Assistant channel.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel (2026). AI referral sources for GA4 (reference list, version 1.0) [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/ai-referral-sources-ga4/ Licence CC BY 4.0.

  • DIY SEO workbook

    What it contains
    A nine-tab spreadsheet for owners who run SEO themselves. It is a template, not research data.
    Window
    Not applicable
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    Placeholder domains only.
    Site names in the file
    Placeholders such as example.com and my own site.
    How it was verified
    Not applicable. It is a blank template.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. DIY SEO workbook [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/diy-seo-workbook/ Licence CC BY 4.0.

  • Migration kit

    • migration-kit.xlsx(XLSX)SHA-256 0302086cad56d6669a7e18fc2d2363587683ddf3d642d62d79620aa24dd3dd92
    What it contains
    A seven-tab spreadsheet: URL map, practice sheet, answer key, baseline, launch-day checks and a 30-60-90 day watch. It is a template, not research data.
    Window
    Not applicable
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    Placeholder domains only.
    Site names in the file
    Placeholders such as example-shop.co.uk.
    How it was verified
    Not applicable. It is a template with a worked example.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. Migration kit [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/website-redesign-seo-checklist/ Licence CC BY 4.0.

  • SEO progress tracker

    What it contains
    A blank review schedule: baseline, first check, early signal check, with columns for impressions, clicks, enquiries, the change made and the next review date.
    Window
    Not applicable
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    No data.
    Site names in the file
    None.
    How it was verified
    Not applicable. It is a blank template.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. SEO progress tracker [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/how-long-does-seo-take/ Licence CC BY 4.0.

  • SEO audit checklist

    What it contains
    A PDF checklist for an SEO audit. It is a document, not research data.
    Window
    Not applicable
    Study page
    None. It is a standalone download.
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    No data.
    Site names in the file
    sunnypatel.co.uk, which is this site, and example domains.
    How it was verified
    Not applicable.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. SEO audit checklist [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/research/data/ Licence CC BY 4.0.

  • Provider brief templates

    What it contains
    Three fill-in briefs to send to every shortlisted provider. They are templates, not research data.
    Window
    Not applicable
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    No data.
    Site names in the file
    None found.
    How it was verified
    Not applicable.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. Provider brief templates [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/best-seo-companies-uk/ Licence CC BY 4.0.

  • Worked examples (fictional)

    What it contains
    A content brief extract and a technical audit finding. Both say at the top that they are fictional worked examples.
    Window
    Not applicable
    Study page
    None. It is a standalone download.
    Sample
    Not applicable
    Exclusions
    None
    Anonymisation rule
    They use a fictional business, and not a real client.
    Site names in the file
    Only the fictional business named in the example.
    How it was verified
    Not applicable. They are not real results.
    Licence
    CC BY 4.0. Use it, share it and adapt it with credit. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. Worked examples (fictional) [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/research/data/ Licence CC BY 4.0.

  • AI referral study verification script

    What it contains
    A Python script with no network calls and no extra packages. It recomputes the study's headline table from the public JSON.
    Window
    Runs on the edition 2 window, 24 June to 21 September 2026
    Sample
    The 108-property frame in the JSON
    Exclusions
    It does not pull fresh GA4 data, rerun source classification or verify the original extraction.
    Anonymisation rule
    It reads only the anonymised JSON.
    Site names in the file
    None found.
    How it was verified
    It checks arithmetic only. Run it as: python verify-ai-referral-study-v2.py ai-referral-traffic-study-2026-09-21-v2.json
    Licence
    MIT. Use it, share it and adapt it, and keep the licence name with it. Licence text
    DOI
    No DOI yet
    Suggested citation
    Sunny Patel. AI referral study verification script [Resource]. sunnypatel.co.uk. https://sunnypatel.co.uk/blog/ai-referral-traffic-study/ Licence MIT.

When do the files get a DOI?

Each entry says No DOI yet. I plan to add a Zenodo record, and I will put the DOI here when it exists.

Until then, cite the study page and the date. The methodology explains how each file was built.

Want this kind of check on your own site? The free 20-minute diagnosis is at /contact/.