UK Footwear AI Visibility 2026, Part 2: The Playbook

Part 1 ranked 10 UK shoe brands by how often ChatGPT, Claude, Perplexity and Gemini name them. Loake led at 66% Share of Voice; Pavers placed last at 9.5%, a 7× gap between first and last. If you have not read it, start with the ranking - this part assumes you know where your brand landed.

A ranking shows the scoreboard. It does not say what to do about it. Part 2 is the action plan: one playbook for each of the three tiers the ranking produced, with the interventions that move Share of Voice and the timeline each tier should realistically expect.

The footwear gap is not unique to footwear

Before the tier plans, one finding gives them weight. The pattern in the footwear ranking - a few well-covered brands far ahead of a long, low tail - is not specific to shoes.

Alongside the UK Footwear study, GEOlikeaPro maintains a standing brand corpus: 54 consumer brands across six categories and eighteen countries, re-audited every month. Of the 34 brands audited so far, the average Share of Voice is 16.5%. Two-thirds sit at 15% or below, and four are not cited once. Eight of the 34 clear 30%; the other twenty-six do not.

The corpus runs broad category queries rather than the UK-shoe queries from Part 1, so its numbers are not a like-for-like leaderboard. They are a pattern check, and the pattern holds in every category measured: a short head, a long tail.

One corpus result reframes the problem. Sentiment - how positively a model describes a brand it does mention - stayed between 70 and 90 for every brand cited at all. Manufactum, for example, records a 3% Share of Voice and a 90% sentiment score. Models are not describing low-ranked brands badly. They are not describing them at all. The constraint is citation frequency, not reputation.

Tier 1: Defend the lead - Loake, Grenson, Joseph Cheaney, Hotter Shoes

For the top four, the risk is not a competitor. It is complacency and unmonitored technical decay. The editorial coverage that built the lead - decades of it in the Financial Times, GQ, Permanent Style and similar titles - has a half-life: models retrain, and newer brands earn newer coverage. A lead is held by being re-won each quarter.

1. Measure monthly. Re-run the category queries each month with GEOlikeaPro's SOV Dashboard or a logged prompt set. A 10-point drop is an early warning that only surfaces if someone is watching for it.
2. Confirm AI crawlers are allowed. Some heritage brands block GPTBot, ClaudeBot, PerplexityBot and Google-Extended in robots.txt without realising it. If a crawler cannot read the site, the brand's own pages cannot reinforce what models already believe. Check yourdomain.com/robots.txt and confirm all four are permitted.
3. Keep earning coverage. One feature per quarter in a publication models cite - Permanent Style, The Rake, Esquire - is maintenance, not growth. Stopping lets the half-life erode the lead.
4. Cover the weak queries. Most leaders win some of the 10 Part 1 queries and trail on others. Identify the two weakest and direct a content push there rather than reinforcing queries already won.
5. Ship Product and FAQ schema. It is table stakes; a leader without structured data leaves an avoidable gap. Validate the markup with Google's Rich Results Test (search.google.com/test/rich-results).

Tier 2: Close the gap in 90 days - Dune London, Vivobarefoot, Sole Bliss

The 90-day window applies to this tier. These brands are visible, at 31% to 36% Share of Voice, but not on the shortlist: models know they exist and do not reach for them first. It is the most improvable position in the ranking. The sequence below runs in order.

1. Audit the technical floor first, week 1. Crawl the site with Screaming Frog (screamingfrog.co.uk, free to 500 URLs). Confirm robots.txt allows the four AI crawlers, sitemap.xml resolves, and Organization schema is present.
2. Add FAQ schema to product and category pages, weeks 1 to 2. Use real shopper questions, answered in plain language and marked up as FAQPage. The GEOlikeaPro FAQ Generator produces these from a URL. Models lift such answers close to verbatim.
3. Reverse-engineer three lost queries, weeks 2 to 3. Take three Part 1 queries where a Tier 1 brand wins and ask each model to cite its sources. Those URLs are the target publication list.
4. Earn three to five editorial mentions, weeks 3 to 12. This is the slow lever and the one that actually moves Share of Voice. Reporter-query platforms Featured (featured.com) and Qwoted (qwoted.com) provide a route in. Vivobarefoot, for instance, has an underused angle with barefoot and foot-health journalists.
5. Re-measure at day 90. A realistic result is a single-digit to low-double-digit gain - a move from 32% to 42% is a genuine win - not a jump into the top four.

Tier 3: The hard climb - Schuh, Wide Fit Shoes, Pavers

This tier is a two-quarter project, not a 90-day one, and two of the three brands face a structural problem schema cannot fix. Schuh and Wide Fit Shoes are retailers. Asked a query like "where to buy quality British shoes," an AI model leads with brand recommendations - which maker to buy - and surfaces retailers only as sellers beneath a chosen product. A multi-brand store is rarely the recommendation itself, and is not guaranteed to appear even as a seller. That is a structural position, not a markup problem.

1. Retailers: build editorial, not inventory pages. A filtered grid of products is not citable. A buying guide such as "how to choose wide-fit shoes," written by a named in-house expert with a byline, is.
2. Become a resolvable entity. Models cite entities they can identify. A maintained Wikipedia page, consistent name and address details across the web, and Organization schema with a sameAs block all help a model place the brand.
3. Pavers: pursue the comfort angle. Hotter Shoes reached 4th at 51.5%, the only non-heritage brand in the top tier, on years of comfort and value review coverage. That route is open and proven; consumer-review and older-shopper publications fit it better than menswear titles.
4. Publish original data. A retailer is rarely written about like a heritage brand, so give journalists a reason to cite it: proprietary sales or fit data, a small survey, a category trend. Coverage follows useful data.
5. Set the timeline before starting. Plan for two quarters minimum. Teams that expect results in 90 days tend to abandon the editorial work before it compounds.

What every tier shares

Across all three playbooks the lever is the same: mention density in the sources models trust. Schema is the floor - it makes a brand machine-readable and eligible for citation - and the mid-market e-commerce audit caps its effect at under 10 points of Share of Voice. Editorial coverage, earned in publications models cite, is the ceiling, and it accounts for most of the score.

Heritage brands lead Part 1 not because their markup is better, but because they have been written about for decades. That history cannot be bought. This quarter's coverage can be earned.

Part 3 will break down where the four models disagree: ChatGPT, Claude, Perplexity and Gemini do not return the same brands for the same query, and that divergence is its own opportunity.