Research plans

Updated 18 November 2026. These are the studies the Jacob Whitmore Index will run next. Each plan is written down before any data is collected, so the method can't drift to fit the result. Where a plan has a first reading, it links to it. Anything without a link has not been measured.

Zero-click: what Google shows above the first link

QuestionFor everyday UK questions, how often does Google show an answer above the first ordinary link, and how far down the page does that link start?
StatusPilot published 18 Nov 2026: 40 of 40 questions. Index row JWI-11.
SampleThe same 40 questions every reading, in four groups of 10: quick facts, how-to, health, money. Changed 18 Nov 2026 from a planned 60: buying and local questions depend on where the searcher is, so they would not compare month to month from one location.
CollectionEach question searched once on google.co.uk (hl=en-GB, gl=uk, pws=0), desktop, with at least 20 seconds between searches. A published browser script records which blocks are visible and where they start.
MeasuresAI Overview shown (yes/no), Google answer box shown, adverts, map pack, People also ask, pixel depth of the first ordinary result, the domain holding it.
LimitsOne location per reading. Google varies by place, time and account. Shows what was on the page, not what people clicked.
RepeatMonthly, around the 18th.
Why it mattersZero-click is widely discussed and usually quoted from US clickstream panels. A dated, open, UK-specific reading with the questions listed is something a reporter can check and cite.

Which UK sites Google shows for everyday questions

QuestionWhich domains hold the first ordinary result and the top ten for everyday UK questions, and does that change month to month?
StatusPlan. Uses the zero-click panel, so no separate collection is needed. The pilot already shows public bodies (gov.uk, NHS, Bank of England, MoneyHelper) holding the first link on 18 of 40.
SampleThe same 60-question panel.
CollectionAdd the top ten ordinary result domains to each zero-click reading.
MeasuresShare of first results and top-ten slots by domain and by type (public body, publisher, retailer, forum, video).
Limits60 questions is not the whole of UK search. Ranking tools with millions of keywords cover more; this panel is small but fully published.
RepeatMonthly, with the zero-click reading.
Why it mattersShows whether official sources and forums are taking ground from independent publishers, with every query visible.

Which UK businesses AI tools recommend

QuestionWhen people ask AI tools for a recommendation in a UK sector, which businesses get named, and do the answers agree across tools and over time?
StatusRetail pilot published 15 Nov 2026 (JWI-10, which UK retailers AI tools name). Nine more sectors planned in UK businesses in AI answers.
SampleTen sectors, ten businesses each, picked from published market data before any answer is read. Fixed question set per sector.
CollectionEach question asked in ChatGPT, Google AI Mode, Gemini and Claude, logged out where possible, answers saved in full.
MeasuresNamed or not, position in the answer, whether a link was given, agreement between tools, change between readings.
LimitsAI answers vary run to run. Each question is asked more than once and the spread is reported, not one answer.
RepeatQuarterly.
Why it mattersBusinesses want to know if AI tools mention them. Many published numbers come from vendors selling trackers. This one publishes the questions and the raw answers.

Do AI answers cite sources that support them?

QuestionWhen an AI search tool cites a source, does that source say what the answer claims?
StatusPlan. Builds on the statistics traced to source method.
Sample50 fixed questions on search and marketing, asked in four AI tools. Every cited link opened and checked.
CollectionSave the answer and its citations. For each claim with a citation, open the source and record what it actually says.
MeasuresShare of citations that support the claim, partly support it, say something else, or fail to load. Results by tool.
LimitsJudging support involves reading, so each call is written down with a quote from the source. A second reader checks a sample.
RepeatTwice a year.
Why it mattersPeople act on cited answers because the link looks like proof. I have not found a UK check that publishes whether the proof holds.

UK Website Quality Index

QuestionOn checks anyone can repeat, how well do the UK's best-known business websites do: speed, accessibility, crawler access and the basics of being found?
StatusPlan. Not started.
SampleThe 100 businesses in the AI brand panel, so quality can be set against AI visibility.
CollectionPublic tools only: Google's PageSpeed Insights and Chrome UX Report field data where published, an automated accessibility scan of the homepage, robots.txt, sitemap and structured data checks.
MeasuresCore Web Vitals pass or fail from field data, count of automated accessibility errors, AI crawler rules, presence of a sitemap and valid structured data.
LimitsAutomated accessibility scans catch a minority of real problems; the count is a floor, not a grade. Homepages only in the first reading.
RepeatTwice a year.
Why it mattersLets the Index test a claim the industry repeats without evidence: that better-built sites get mentioned more by AI tools.

Digital marketing claims, checked

QuestionHow many widely repeated statistics and claims about SEO and AI search trace back to a source that supports them?
StatusStarted. AI search statistics, traced to the source (29 Oct 2026, rerun due 29 Jan 2027) and BrightonSEO claims are the first two readings.
SampleThe most-shared statistics in a set of conference talks, reports and LinkedIn posts, picked before checking.
CollectionFollow each claim back to its first source. Record the chain.
MeasuresSupported, changed in the retelling, no source found, or contradicted.
LimitsThe sample is what gets shared, not everything said. No claims about intent; only whether the source supports the number.
RepeatQuarterly.
Why it mattersA short list of claims with their real sources is the kind of page other writers link to when they correct a number.

The annual report

QuestionWhat changed in UK search and discovery this year, measured rather than predicted?
StatusFirst edition published 18 Nov 2026: The Future of Search Report 2026. One annual report, not several.
SampleEvery Index reading from the year.
CollectionNo new collection. Pulls the year's readings together.
MeasuresYear-on-year change for every Index row that has two readings.
LimitsOnly covers what the Index measured.
RepeatOnce a year, January.
Why it mattersGives journalists one page to cite at the start of the year, with links back to every dataset.

Rules every plan follows

  • The sample and the questions are fixed and published before collection.
  • Every reading has a date, a CSV and the script or steps to reproduce it.
  • Pilots are called pilots. A result is not reported as a share of the UK unless the sample supports it.
  • If a plan changes, the change and the date stay visible on this page.
  • No surveys, panels or numbers from tools that will not show their method.

See also: how I work, all datasets and how to reproduce a study.