Is AI traffic actually worth optimising for yet?

It's small but compounding fast, on Shopify's own numbers: AI chatbot referral sessions grew more than 8x year-over-year and AI-referred orders nearly 13x (Q1 2026). Quality is the real story — those sessions convert at nearly 50% higher rates than organic search and carry 14% higher average order values. Small base, steep curve, better buyers: that's exactly when being early pays.

How long until AI assistants pick up my changes?

Assistants with live browsing (ChatGPT with search, Perplexity) can reflect changes within days of recrawling. Recommendations grounded in trained knowledge or cached indexes move slower — weeks to months. That lag is the argument for starting now: authority signals compound, and AIs cite what is already being cited.

How to get ChatGPT to recommend your store

Q: Does blocking GPTBot hurt my sales?

Blocking AI crawlers makes your store invisible to the assistants shoppers increasingly ask for recommendations — you're opting out of a channel where, on Shopify's published data, referred sessions convert at nearly 50% higher rates than organic search. The trade-off is yours to make (some merchants object to AI training on their content), but make it knowingly: robots.txt disallow means AI assistants recommend your competitors instead.

Q: Does llms.txt actually work?

Honestly: it's an emerging standard with uneven adoption — some AI fetchers use it, others don't yet. It costs almost nothing to publish, gives assistants a curated map of what to cite, and adoption is growing. Treat it as a low-cost, high-upside layer on top of the fundamentals (crawler access, schema, answer-first content) — not a substitute for them.

AI assistants recommend stores they can read, verify, and quote. To become recommendable: allow AI crawlers (GPTBot, ClaudeBot, PerplexityBot) in robots.txt; publish clean structured data (Product, Offer, Organization schema) so your prices and stock are machine-readable; write answer-first product and FAQ content an AI can lift verbatim; add an llms.txt map; and build the review and citation signals AIs use as trust proxies. The channel is small but steep: on Shopify's own published data, AI chatbot referral sessions grew more than 8× year-over-year and AI-referred orders nearly 13× (Q1 2026) — and those sessions convert at nearly 50% higher rates than organic search. Early movers win outsized share.

No one outside OpenAI knows ChatGPT's ranking internals, and anyone selling you "guaranteed AI rankings" is selling snake oil. What is knowable is the mechanism — how an assistant goes from a shopper's question to a store recommendation — and every step of that mechanism is something you can be good or bad at. This guide covers the mechanism, then the six changes ranked by impact.

The Bilby scan screen mid-visit on the demo store — The scan screen mid-visit on our labelled demo store, Fernwick & Co.: a screenshot per step as it lands, then the seven things it checks read the recording. Run it on your store: free, about a minute, no signup, we never touch your store. Run the free scan →

How AI assistants actually pick products

When someone asks "best merino tee that ships to Australia", an assistant with browsing does roughly three things:

Retrieval. It searches and fetches candidate pages — but only from sites its crawlers can access. Blocked at robots.txt means not in the candidate set at all.
Trust filtering. It weighs what it already "knows": review volume and sentiment, citations from articles and communities, whether your brand resolves to a consistent entity across the web. AIs cite what is already being cited.
Extraction. It reads the fetched pages and pulls out claims it can quote confidently — price, materials, shipping, returns. Clean schema and answer-first copy get extracted; a wall of lifestyle prose with the facts buried in an image does not.

// How a product ends up in an AI answer — and where stores fall out of the running.

On Shopify's published data, AI chatbot referral sessions grew more than 8× year-over-year and AI-referred orders nearly 13× in Q1 2026 — and AI-referred sessions convert at nearly 50% higher rates than organic search.¶

The six changes, ranked by impact

1. Let the crawlers in (robots.txt)

Table stakes. Check yourstore.com/robots.txt for disallows against GPTBot, ClaudeBot, PerplexityBot, Google-Extended. On Shopify you customise robots.txt via the robots.txt.liquid template. Blocked crawler = out of the candidate set entirely, whatever else you do. Bilby's Answer check reads those robots.txt lines on every visit and flags a blocked AI crawler as a finding.

2. Clean structured data (Product, Offer, Organization)

Schema is the machine-readable version of your catalogue: exact price, currency, availability, ratings. Most themes emit some Product schema; verify it's valid and complete rather than assuming. An assistant that can read your price with confidence can quote it; one that can't moves on to a store it can. The Answer check also checks that Product and Organization schema are present.

3. Answer-first product and FAQ copy

Write the first two sentences of a product page so they fully answer "what is this and who is it for" — extractable, quotable, no scroll needed. Add visible FAQs answering the questions shoppers actually ask (sizing, shipping, returns) with FAQPage schema. This is the same discipline that wins featured snippets; AIs reward it more.

4. llms.txt

A curated map of what matters on your site, at /llms.txt. Emerging standard, uneven adoption, near-zero cost — the honest full picture (and the Shopify root-file workaround) is in our llms.txt guide, and our free generator builds the file in a minute; the Answer check checks the file exists on every visit.

5. Reviews and citations (the trust proxies)

Assistants lean on external evidence: review volume and sentiment, mentions in "best X" articles, Reddit and community threads. You can't fake this (please don't — fake reviews are the fastest way to poison the well), but you can earn it deliberately: ask for reviews systematically, pitch genuinely useful data to writers in your niche, be the source others cite.

6. Entity consistency

Same brand name, same description, same facts — across your site, socials, app listings and profiles. Organization schema with sameAs links ties it together. Assistants resolve brands into entities; inconsistency reads as uncertainty, and uncertain things don't get recommended. The same goes for the product facts themselves — we checked the product data on 409 live Shopify stores to see what is usually missing.

What NOT to do

Schema spam — marking up ratings you don't have, fake aggregate scores. Assistants cross-check; getting caught costs more than the markup earned.
Fake reviews — the trust proxies exist because they're hard to fake; bought reviews are detectable and radioactive.
"Guaranteed AI ranking" services — nobody can guarantee placement in a system whose internals they don't control. The fundamentals above are the whole legitimate playbook.

More than half of AI-referred shopping sessions land directly on a product page, versus about 20% for organic search — AI assistants send buyers, not browsers, which is why those orders carry 14% higher average order values on Shopify's data.¶

Where do you stand today?

// 6-POINT SELF-AUDIT — TICK WHAT'S TRUE OF YOUR STORE

My robots.txt allows GPTBot, ClaudeBot and PerplexityBot (I've actually checked) My product pages have valid Product + Offer schema with live price and availability My product pages answer "what is this and who is it for" in the first two sentences I publish an llms.txt (or know exactly why I don't) I have a systematic way of earning reviews, and external pages cite my store My brand name, description and facts are identical across site, socials and listings

Tick the boxes to see your score.

Check your store's AI visibility automatically. The free AI Store Check reads your robots.txt per AI crawler, validates your schema, looks for llms.txt and scores your extractability — in about a minute. On Bilby Pro, the Answer check goes further: monthly it asks ChatGPT, Perplexity and Claude the same five buyer questions your shoppers ask, stores the raw responses, and reports whether your store is recommended — and who was recommended instead. And if you'd rather the content side ran itself, Navaal: AI SEO & GEO writes and maintains answer-first, schema-complete product content across your whole catalogue.

Frequently asked questions

Does blocking GPTBot hurt my sales?

It makes you invisible to assistants shoppers increasingly ask for recommendations — a channel converting at nearly 50% higher rates than organic search on Shopify's data. Some merchants block anyway on principle; make the trade-off knowingly, because the practical effect is that AIs recommend your competitors.

Is AI traffic worth optimising for yet?

Small base, steep curve, better buyers: sessions up 8×+ YoY, orders up ~13× (Shopify, Q1 2026) — converting ~50% better with 14% higher order values. That combination is exactly when early positioning pays.

Does llms.txt actually work?

Emerging standard, uneven adoption, near-zero cost, growing. A useful layer on top of the fundamentals — never a substitute for crawler access, schema and answer-first content.

How long until results?

Assistants with live browsing can reflect changes within days of a recrawl; trained-knowledge recommendations move over weeks to months. Authority compounds — which is the argument for starting before your competitors do.

Keep reading: The complete llms.txt guide for Shopify · What is GEO? · Free AI Store Check

// Related field notes

Keep reading

AI search

Is your store
visible to AI?

The free AI Store Check reads your store the way ChatGPT does — crawler access, schema, extractability — and scores it in a minute.

Check my AI visibility →Free store scan How the Answer check works