3 November 2026

Measurement · AI & organic search

The baseline everything after it is judged against.

Three measures, in the order we will report them every month: revenue from non-brand discovery, then non-brand click-through against the position curve, then AI presence and citations. Revenue leads because it is the only one of the three that is an outcome rather than a proxy. Two of the three cannot be completed from the inputs assembled for them, and this document says so rather than estimating around it.

Reading taken: 3 November 2026
Window: 1 Jun 2025 – 30 Sep 2026, frozen
Contractual date: 30 October 2026
Prepared by: Matt Jones

Status

The plan inverted.

This was scheduled on the basis that every input was assembled and it was a writing job. The section expected to be pending is the one that is finished.

Revenue from non-brand discovery

Cannot complete

Organic revenue is absent from the inputs — 0 of 991 rows, across all seven columns.

Non-brand click-through vs the position curve

Cannot complete

Only URL-level average position exists. A CTR curve has to be built at query level.

AI presence and citations

Complete

221 prompts run across Perplexity, ChatGPT and Gemini. 663 answers, zero errors.

Nothing below has been estimated or filled in. Where a figure is not in the inputs, that is stated in place of a figure.

Limits that apply throughout

Amazon is invisible to this measurement

Bibado's Amazon trade appears in no connected system. Every revenue figure here is site revenue only, and Amazon's share of total has been requested but is not in the inputs.

Search Console anonymises a share of clicks

Low-volume queries are withheld, so any brand and non-brand split is an upper bound on non-brand rather than an exact figure.

AI attribution is partial

Assistants that strip the referrer or answer without a click are not counted. The referral figures in 3d are a floor, not a measure of influence.

Small samples carry their variance

Reported at the number rather than left to inference. Several figures below rest on very few rows and say so.

GA4 applies consent modelling

Visitors who decline cookies are modelled rather than observed, which bears hardest on the smallest segments — which is most of section 3.

Revenue from non-brand discovery

£1.76m

All-channel GA4 revenue over the 16-month window

138

URLs earning any revenue, of 991 mapped

866,626

GA4 sessions in the window

57%

Of all revenue on the top two URLs

What is established

Revenue is heavily concentrated. The top two URLs carry 57% of all site revenue in the window, and the Coverall across its two product URLs a further 25%. That concentration is itself a baseline fact worth carrying forward: the commercial exposure of this site sits on four URLs.

GA4 revenue by URL · 16 months to 30 Sep 2026
URL	Type	Revenue
/	Landing	£707,586
/pages/bundles	Landing	£298,039
/products/coverall-weaning-bib-long-sleeves	Product	£272,793
/products/coverall-weaning-bib-long-sleeves-2023	Product	£163,175
/products/3-bib-bundle	Product	£37,354
/collections/weaning-coveralls-baby-bibs	Collection	£36,424
/collections/christmas-sale-all-bundles	Collection	£31,499
/products/new-coveralls-long-and-short-sleeves	Product	£28,162

GA4 organic revenue is entirely absent

All seven organic-revenue columns in the assembled inputs are empty:

GA4 organic revenue | 16 months · Jun 2025–Sep 2026     0 of 991 rows
GA4 organic revenue | Prior 3m · Apr–Jun 2026           0 of 991 rows
GA4 organic revenue | Latest 3m · Jul–Sep 2026          0 of 991 rows
GA4 organic revenue | 3-month change                    0 of 991 rows
GA4 organic revenue | Prior 6m · Oct 2025–Mar 2026      0 of 991 rows
GA4 organic revenue | Latest 6m · Apr–Sep 2026          0 of 991 rows
GA4 organic revenue | 6-month change                    0 of 991 rows

The headline metric of this engagement therefore has no baseline value, and no claim about growth in non-brand revenue can be made or refuted.

There is no brand versus non-brand split available anywhere. That split requires query-level Search Console data. The inputs hold URL-level data only, so non-brand cannot be isolated in this section or the next.

The period comparisons are not usable — do not quote them

The 3-month and 6-month revenue columns are populated for between 4 and 21 of 991 URLs. They produce arithmetic — £19,234 for Oct 2025–Mar 2026 against £20,440 for Apr–Sep 2026, apparently +6.3% — resting on 4 rows and 20 rows, implying £40k across twelve months of a window whose total is £1.76m. Recorded only so that nobody later mistakes it for a trend.

What would close it

A GA4 pull of revenue by session default channel group, organic only, over the same frozen window, at more than one attribution model
A query-level Search Console export, to derive the brand and non-brand split as an upper bound on non-brand
Amazon revenue as a share of total, from Bibado, so the baseline can state what it cannot see rather than implying the site is the whole business

Non-brand click-through against the position curve

The site totals are sound and are usable baseline numbers in their own right. The curve the section calls for is not available from these inputs.

97,208

Clicks over the 16-month window

6.36m

Impressions

1.53%

Site-wide click-through rate

990

Mapped URLs

Why the curve cannot be built

The inputs give average position per URL, not per query. Grouping URLs by their average position produces this:

URL-level position bands · this is not a CTR curve
Position band	URLs	Clicks	Impressions	CTR
1	3	1	3	33.33%
2–3	12	1,512	91,344	1.66%
4–5	39	509	98,257	0.52%
6–10	299	47,867	3,972,122	1.21%
11–20	206	39,619	1,601,955	2.47%
21–50	139	572	63,997	0.89%
51+	23	52	8,785	0.59%

This is not a CTR curve and must not be presented as one

It is non-monotonic. Positions 11–20 appear to out-click positions 6–10 by two to one, and positions 4–5 appear worse than everything below them. The position-1 row rests on three impressions.

The reason is structural, not a data error. A URL's average position across thousands of queries is not its rank for any single one of them. A guide averaging position 15 across 2,000 queries may rank second for the handful that actually convert, and those clicks land in the 11–20 band. Averaging position before measuring click-through destroys the relationship the metric exists to capture.

What would close it

A query-level Search Console export: query, clicks, impressions, average position. CTR by position can only be computed at query level
Brand and non-brand classification applied to those queries. Brand terms sit at position 1 with very high click-through and will dominate any unsegmented curve, which is precisely why the specification asked for non-brand
A named published position curve to measure the ratio against. The specification calls for a ratio to "the published position curve" without naming one, and the ratio is meaningless without knowing its denominator

AI presence and citations

221 questions a UK weaning parent would plausibly ask, put to Perplexity (sonar), ChatGPT (gpt-4.1, web search) and Google Gemini (gemini-3.6-flash, Search grounding), all pinned to the UK. 663 answers, zero errors. 15 competitor brands tracked by name across identical questions, so every comparison is like-for-like.

Bibado named in answers where the mention has to be earned — 33 of 612

3rd

Of the 15 brands measured

41%

On authority questions — the strongest brand measured

On product questions, against ezpz's 24%

The 51 brand-name answers are excluded, and must stay excluded. Bibado scores 100% on them, which measures nothing except that the question contained the word Bibado.

3a · Earned mentions across the 612 category, product and authority answers
Brand	Mentions	Share
ezpz	54	9%
Munchkin	48	8%
Bibado	33	5%
Bumkins	24	4%
Olababy	18	3%
Grabease	15	2%
Tidy Tot	12	2%
Doddl	11	2%
Stokke	10	2%
Mushie	8	1%
Tommee Tippee	6	1%
Nuby	4	1%
Vital Baby	4	1%
Bibetta	1	0%
BabyBjörn	1	0%

Where Bibado wins and loses
Question type	Answers	Bibado	Best competitor
authority	27	41% (11)	ezpz / Munchkin 15%
product	195	7% (14)	ezpz 24% (47), Munchkin 22% (43)
category	390	2% (8)	Bumkins 2% (6)

Category: 2% is not a loss

377 of the 390 category answers named no brand at all. The engines give advice and cite the NHS; they do not recommend products. Of the 13 that did name a brand, Bibado took 8 — more than Bumkins (6), Tidy Tot (4) and ezpz (3). On the subset where any brand is named, Bibado leads.

So the opportunity on category questions is to be the cited source, not the named brand. Bibado's own site was cited in 1 of 390 category answers.

Product: the real, addressable gap

62 of the 71 questions where a competitor was named and Bibado was not are product questions. ezpz appears in 52 of those gaps, Munchkin in 43. These are not head terms like "best weaning bib UK" — they are specific and problem-framed: my baby grabs the rim of their cup instead of the sides and tips it everywhere, what shape is easiest for little hands to control? Each gap is effectively a content brief.

3b · Who the engines trust on weaning — answers citing each domain, of 663
Domain	Answers	Domain	Answers
nhs.uk	66	cdc.gov	26
solidstarts.com	35	raisingchildren.net.au	26
justonenorfolk.nhs.uk	34	bibado.com	26
srnutrition.co.uk	34	myhealth.alberta.ca	24
healthychildren.org	29	mylittleeater.com	23
doddl.com	28	babylist.com	22
nibli.org	28	bibado.co.uk	19

NHS and NHS trust domains dominate and are not realistically influenceable. The reachable targets are the independent publishers — solidstarts, srnutrition, nibli, mylittleeater, babylist. Two findings to carry forward:

doddl.com (28) is cited in more answers than bibado.com (26). A direct competitor's own site is cited marginally more often than Bibado's.
Bibado's citations are split across two domains — bibado.com 26 and bibado.co.uk 19. If that is not a deliberate US and UK separation, it divides the visible authority of one brand across two. The Ahrefs tracking in 3c counts bibado.co.uk only, so that figure may understate Bibado for the same reason.

3c · Ahrefs citation share · United Kingdom, all platforms, snapshot 15 Oct 2026
Brand	Citations	Cited pages	Share
Tommee Tippee	1,665	858	33.6%
Nuby	1,298	327	26.2%
BabyBjörn	795	270	16.1%
Munchkin	605	155	12.2%
Bibado	212	62	4.3%
doddl	157	89	3.2%
Done by Deer	105	86	2.1%
bamboo bamboo	89	29	1.8%
Tidy Tot	19	9	0.4%
Bibetta	5	5	0.1%
Bumkins	2	1	0.0%

The denominator is 4,952 citations across this set of eleven brands — a share of the tracked set, not of the category. Google surfaces (AI Overviews, the AI Overviews keyword index and AI Mode) account for 162 of 212 UK citations, 76%. Bibado earns 3.4 citations per distinct cited page against Tommee Tippee's 1.9 and Nuby's 4.0: a narrow base of 62 pages, each cited harder. The lever is breadth of citable pages, not depth on the ones already working.

3d · AI referral traffic and revenue · GA4 via SEO Gets, all countries, 16 months
Platform	Sessions	Engagement	Key events	Revenue
All identified AI assistants	1,189	41.55%	49	£1,787.73
ChatGPT	1,126	41.39%	47	£1,733.93
Gemini	23	43.48%	1	£37.90
Microsoft Copilot	17	29.41%	1	£15.90
Perplexity	15	60.00%	0	£0.00
Claude	7	57.14%	0	£0.00
DeepSeek	1	0.00%	0	£0.00

0.14%

AI share of all sessions

0.101%

AI share of all revenue

£1.50

Revenue per AI session

94.7%

ChatGPT share of AI sessions

AI referral is immaterial to revenue today. A tenth of one percent. Anyone arguing that AI search is already moving Bibado's numbers is not supported by this data. The case for the work is the trajectory and the cost of being absent later, not present revenue.
ChatGPT is effectively the whole of measured AI referral traffic at 94.7% of sessions, while contributing only 14 of 212 citations. Citations and referrals are near-inverted.
Everything below ChatGPT is too small to trend. Perplexity's 60% engagement rests on 15 sessions and Claude's 57% on 7. Neither should be reported as a finding.

The three measures disagree

This document now holds three separate measures of AI presence. They are not interchangeable, and the competitor rankings they produce are close to contradictory.

Brand	Ahrefs UK citations	Earned mentions
Tommee Tippee	1,665 — 1st	6 — 1%
Nuby	1,298 — 2nd	4 — 1%
BabyBjörn	795 — 3rd	1 — 0%
Munchkin	605 — 4th	48 — 2nd
Bibado	212 — 5th	33 — 3rd
ezpz	not tracked	54 — 1st
Olababy	not tracked	18
Grabease	not tracked	15
Stokke	not tracked	10
Mushie	not tracked	8

The brand that wins AI answers on weaning is not in the tracked competitor set

ezpz leads the prompt measurement at 9% and does not appear in the Ahrefs competitor list at all. Nor do Olababy, Grabease, Stokke or Mushie. Meanwhile the top three by Ahrefs citations are close to invisible in actual weaning answers. The competitor set used for Ahrefs tracking needs rebuilding around the brands that surface in answers.

The divergence is explained, not a data fault. Ahrefs counts citations of a domain across its AI keyword index, largely AI Overviews, which supplied 115 of Bibado's 212 UK citations. Tommee Tippee has 858 cited pages spanning bottles, monitors and sterilisers, so its count reflects breadth across all of baby rather than weaning authority. The prompt test counts whether a brand is named in answers to 221 weaning questions. For a weaning specialist the prompt measurement is the more relevant of the two and should be the headline; the Ahrefs series stays useful as a domain-level trend, but not as a competitive ranking.

Taken together, citations, named mentions and referral traffic each put a different brand in front. Google surfaces drive citations, ChatGPT drives 94.7% of referrals while earning 14 of 212 citations, and ezpz wins named mentions. Any future report that collapses these into one AI visibility number will be wrong.

Limits specific to this measurement

Single reading. 3 November 2026 only. Nothing here is a trend, an improvement or a decline.
Not comparable to the September 2026 snapshot. An earlier 50-question snapshot recorded 49% presence. It shares one question with this set and was weighted completely differently. The two measure different things and must not be placed side by side. The original 50 were not preserved as a fixed subset, so no like-for-like September comparison is available.
Gemini's citations are unreadable. 96 of 221 Gemini answers cite Google's grounding redirect wrapper rather than a named publisher, so the citation table in 3b effectively covers Perplexity and ChatGPT only. Brand presence figures are unaffected, since they read the answer text rather than the links.
The authority figure rests on 27 answers. 41% is directional, not solid.
Presence is not recommendation. The measure is whether a brand name appears anywhere in the answer. A passing mention counts the same as a recommendation.
The question set is not blind. It includes Bibado product names (Dippit, Sporkit, Sippit, Big Sipper), so it was written with knowledge of the range. Acceptable for tracking, but it has to be stated if these figures go to the client.
Not a clean pre-change reading. The prompt task specified that it must precede any change on site, but broken internal links, Recipe schema, indexable test pages and the HTTP stylesheet reference had already shipped by 3 November. The Ahrefs snapshot of 15 October predates most of that work and is the cleaner pre-change reference.

Phrasings this data does not support

Recorded so that later reports do not drift into them.

✕Bibado leads share of voice or the most-mentioned brand — only true with the brand-name questions left in, which makes it circular
✕13% AI visibility — that figure includes the brand questions. It is 5%
✕Any comparison with 49%, or any framing as a change since September
✕AI doesn't know Bibado — it does. Brand questions scored 100%
✕Bibado is absent from category questions without the 377-of-390 context. Stated bare it reads as a failure, and it isn't one

Every answer, as it came back

All 663 answers behind the figures above. Filter by engine, question type or whether Bibado was named, and open any row for the answer itself and the domains it cited. The percentages in 3a are simply counts over this table.

Engine

Question type

Bibado

Search the question or answer

#	Question	Engine	Type	Bibado	Brands named in the answer

What this baseline supports

Can be stated and defended today

Site revenue, and its concentration on four URLs
Site-wide clicks, impressions and CTR over a frozen window
Bibado named in 5% of earned AI answers, third behind ezpz and Munchkin
41% on authority questions — the strongest brand measured
The 7%-against-24% product gap, with 62 named content briefs behind it
That 377 of 390 category answers name no brand, so the play there is citation rather than mention
Ahrefs citation share by platform at a dated snapshot
AI referral sessions, engagement and revenue — and that AI referral is immaterial to revenue today

Cannot be stated until the gaps close

Revenue from non-brand discovery — the engagement's headline metric
Any brand versus non-brand split, anywhere
Non-brand click-through as a ratio to a position curve
Any revenue or click-through trend across the comparison periods
What share of Bibado's total trade this measurement can see
Any change in AI presence over time — 3 November is a first reading, and September is not comparable to it

Immediate asks

#	Ask	Owner	Unblocks
1	GA4 organic revenue by session default channel group, frozen window, multiple attribution models	Searchflex	Section 1
2	Query-level Search Console export: query, clicks, impressions, position	Searchflex	Sections 1 & 2
3	Agree and name the published position curve to measure against	Searchflex	Section 2
4	Amazon revenue as a share of total	Bibado — Dan	Limit 1
5	Rebuild the Ahrefs competitor set around brands that actually surface — add ezpz, Olababy, Grabease, Stokke, Mushie	Searchflex	3c as a ranking
6	Decide whether bibado.com and bibado.co.uk is a deliberate split; add bibado.com to tracking either way	Bibado + Searchflex	3b & 3c
7	Freeze the 221 prompts as the tracked set from this reading forward, wording unchanged	Searchflex	Any future trend

Items 1 and 2 are the whole of the difference between a complete baseline and this one. Both are data pulls against systems already connected, not client dependencies.

Item 7 matters more than it looks. The September set was not preserved, which is why no change measurement exists today. If the 221 are not frozen now, the same thing happens again at the next reading.