VC & Startups/Nov 21 edition/Reporting & analysis
FundingEcosystems

FundingRounds, leads & terms

Wilgot raises $3M seed led by Industrifonden to optimise product data for AI shopping assistants

Stockholm startup Wilgot has raised a $3 million seed round led by Industrifonden. It builds software that rewrites brands' product pages, feeds and structured data so AI assistants such as ChatGPT recommend their products, rather than just tracking mentions.

From the Tech.eu article of November 19, 2026 and cited sources: The SEK 30 million seed round was led by Industrifonden, with Luminar Ventures, Foundry Ventures, Ventures Together, Inovexus and founder angels from Voyado and inriver taking part. [1] [5] [6] [7]

SOURCES[1] Tech.eu: Wilgot raises $3M in seed funding to help brands… tech.eu[5] Wilgot: Wilgot raises $3M to make products work for both… wilgot.ai[6] ecommercenews: Wilgot raises USD $3 million to boost AI… ecommercenews.com.au

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
The SEK 30 million seed round was led by Industrifonden, with Luminar Ventures, Foundry Ventures, Ventures Together, Inovexus and founder angels from Voyado and inriver taking part. [1] [5] [6] [7]
Original source
Tech.eu article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
Studies
arXiv 2311.09735 (Nov 2023) [2] date from arXiv ID
When
November 19, 2026 article published
Evidence
7 sources across 7 sites

Reach: not measured yet

Illustration from Tech.eu: Wilgot raises $3M seed led by Industrifonden to optimise product data for AI shopping assistants
Image: Tech.eu — Original article ↗
THE CORE IDEAS4 TAKEAWAYS
01

The SEK 30 million seed round was led by Industrifonden, with Luminar Ventures, Foundry Ventures, Ventures Together, Inovexus and founder angels from Voyado and inriver taking part. Industrifonden's Andreas Söneby and Foundry's Anders Nygren join the board. The valuation was not disclosed. [1] [5] [6] [7]

02

Wilgot calls its approach Agentic Product Optimisation. It takes a brand's own search, conversion, review, support-ticket and returns data and produces product content in more than 100 languages. That content is then refined in a loop based on what gets clicked, recommended and bought. The reward design and the base model behind it have not been documented. [1] [3] [5]

03

Wilgot reports a 20% rise in net sales on optimised Daniel Wellington product pages. It also says 7% of Myrqvist customers arrive through ChatGPT. Both figures come from the company and have not been independently evaluated. [1] [5]

04

OpenAI's Instant Checkout and its Agentic Commerce Protocol let assistants complete purchases, not just recommend products. That raises the value of clean, machine-readable product data. Earlier research on generative engine optimisation found that visibility gains vary widely by domain. [2] [4]

WHY IT MATTERS

What is established: the funding round and investors. What is not: results, which are unpublished.

Read the full assessment

As assistants add checkout, structured product data may increasingly drive sales, but gains depend on opaque, changing ranking systems and may differ by product category.

Executive brief

Wilgot is a Stockholm startup founded in 2024. It has raised a $3 million (SEK 30 million) seed round led by the Swedish venture investor Industrifonden. The pitch is that brands should change their product data so AI assistants recommend them, rather than only tracking how often they get mentioned. Wilgot calls this "Agentic Product Optimisation" (APO). Its headline result is a 20% rise in net sales on optimised Daniel Wellington product pages, but the company reported that figure itself and no outside party has checked it (Tech.eu). The valuation was not disclosed.

What changed and event timeline

  1. GEO research paper published

    Princeton-affiliated researchers coined "generative engine optimization." They introduced a benchmark called GEO-bench and reported visibility gains of up to 40% in AI-generated answers ().

  2. Wilgot founded

    Erik Wikander (CEO, formerly CMO at Lendify), Dr Mhairi Dunion (AI Scientist, PhD in deep reinforcement learning) and Oleg Danylenko (CTO, ex-Klarna and IKEA) started the company in Stockholm ().

  3. ChatGPT adds checkout

    OpenAI launched Instant Checkout and the open Agentic Commerce Protocol, built with Stripe. At launch it covered single-item purchases from US Etsy sellers ().

  4. $3M seed announced

    Luminar Ventures, Foundry Ventures, Ventures Together and Inovexus joined, along with founder angels from Voyado and inriver (;).

  5. Board seats

    Andreas Söneby (Industrifonden) and Anders Nygren (Foundry Ventures) join Wilgot's board ().

Capabilities and access

  • What it is: a software platform for brands. No model, version or public API has been disclosed.
  • Inputs: the brand's own data, including search queries, click-through and conversion data, reviews, service tickets and returns (Wilgot).
  • Outputs: rewritten product pages, product feeds, FAQs and structured data, in 100+ languages.
Read the full section
  • What it is: a software platform for brands. No model, version or public API has been disclosed.
  • Inputs: the brand's own data, including search queries, click-through and conversion data, reviews, service tickets and returns (Wilgot).
  • Outputs: rewritten product pages, product feeds, FAQs and structured data, in 100+ languages.
  • Throughput: the company says it processes 100,000 products per minute (ecommercenews).
  • Named customers: Daniel Wellington, J.Lindeberg, Myrqvist and Signify.

Technical analysis for researchers and developers

  • How it works, per the company: an optimisation loop. Tech.eu describes this as reinforcement learning (Tech.eu).
  • Undocumented: the reward design, the policy architecture, which base LLM generates the content, and how each assistant's recommendations are measured.
  • Hard problem: the reward signal is noisy and comes from systems Wilgot can't see inside (ChatGPT, Gemini, Perplexity).
Read the full section
  • How it works, per the company: an optimisation loop. Product content is generated, then refined based on what "gets clicked, recommended and bought" (Wilgot). Tech.eu describes this as reinforcement learning (Tech.eu).
  • Undocumented: the reward design, the policy architecture, which base LLM generates the content, and how each assistant's recommendations are measured.
  • Hard problem: the reward signal is noisy and comes from systems Wilgot can't see inside (ChatGPT, Gemini, Perplexity). Those systems change their retrieval and ranking without notice, so improvements may not hold over time.
  • Reproducibility: there is no paper, benchmark or public evaluation. GEO-bench (arXiv) is the closest open benchmark, but it tests web answers, not product recommendations.

Claims and evidence

  • Round size ($3M / SEK 30M), lead investor, co-investors and board seats: announced by the company (Wilgot).
  • Valuation: not disclosed.
  • Ventures Together: this co-investor published its own announcement of the round (Ventures Together).
Read the full section
  • Round size ($3M / SEK 30M), lead investor, co-investors and board seats: announced by the company (Wilgot). Tech.eu, ArcticToday and ecommercenews repeat the same figures (ArcticToday). All coverage appears to come from the same press release. There is no filing or unnamed-source reporting to confirm it.
  • Valuation: not disclosed.
  • Ventures Together: this co-investor published its own announcement of the round (Ventures Together).
  • 20% net sales lift (Daniel Wellington) and 7% of Myrqvist customers arriving via ChatGPT: both self-reported. Wilgot has not published a baseline, time window, control group or attribution method.
  • Industrifonden's comments: Söneby says commerce is moving "from clicks to conversations" (ecommercenews). He is speaking about his own fund's lead investment.

Context and prior work

  • Research roots: Aggarwal et al. (KDD 2024) treated AI-answer visibility as a black-box optimisation problem and found that results vary by domain (arXiv).
  • Platform shift: OpenAI's Agentic Commerce Protocol moved assistants from recommending products to completing purchases. That makes clean, machine-readable product data worth more (OpenAI).
  • Positioning: Wilgot sets itself against visibility-tracking tools.
Read the full section
  • Research roots: Aggarwal et al. (KDD 2024) treated AI-answer visibility as a black-box optimisation problem and found that results vary by domain (arXiv).
  • Platform shift: OpenAI's Agentic Commerce Protocol moved assistants from recommending products to completing purchases. That makes clean, machine-readable product data worth more (OpenAI).
  • Positioning: Wilgot sets itself against visibility-tracking tools. It argues brands spend too much effort measuring mentions and too little giving AI a reason to recommend them (ArcticToday).

Limitations, safety and contested findings

  • Results: no independent evaluation of Wilgot's results exists in the reviewed sources.
  • Domain dependence: the GEO paper found that optimisation gains differ widely across domains (arXiv).
  • Unclear attribution: a 20% sales lift on optimised pages could also come from ordinary SEO or conversion-rate effects.
Read the full section
  • Results: no independent evaluation of Wilgot's results exists in the reviewed sources.
  • Domain dependence: the GEO paper found that optimisation gains differ widely across domains (arXiv). A lift in one category may not carry over to another.
  • Unclear attribution: a 20% sales lift on optimised pages could also come from ordinary SEO or conversion-rate effects. The sources do not separate these from gains driven by AI assistants.
  • Platform dependence: Instant Checkout launched only in the US and with limited merchants (OpenAI), so how much traffic AI commerce brings depends on each platform's rollout.

Business and practitioner implications

  • For brands: the data Wilgot uses (reviews, returns, support tickets) is data brands already own.
  • For evaluating vendors: ask for holdout tests, per-assistant attribution and a time window before accepting sales-lift claims.
  • For developers: structured data and machine-readable feeds compatible with the Agentic Commerce Protocol are becoming baseline requirements for e-commerce (OpenAI).
Read the full section
  • For brands: the data Wilgot uses (reviews, returns, support tickets) is data brands already own. Turning it into structured attributes such as intended user and fit is useful regardless of which vendor does it.
  • For evaluating vendors: ask for holdout tests, per-assistant attribution and a time window before accepting sales-lift claims.
  • For developers: structured data and machine-readable feeds compatible with the Agentic Commerce Protocol are becoming baseline requirements for e-commerce (OpenAI).
  • For investors: angels who founded Nordic commerce-software companies (Voyado, inriver) point to a likely channel strategy of partnering with existing commerce platforms.
FOLLOW THE EVIDENCE

The source trail.

Sources (7)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.