The old path to purchase was simple: a shopper searched Google or Skroutz, scanned ten results and picked one. Now a growing share of buyers start by asking ChatGPT, Gemini or Perplexity, or they read the AI Overview at the top of Google before scrolling anywhere. What they get back is not ten blue links. It is two or three recommendations, with shop names and product names.
If your shop is not in that short answer, you do not exist for that customer. Here is what you can actually do about it.
How an AI "decides" which shop to recommend
There is no single magic ranking factor, but the patterns are fairly clear:
- Live retrieval. For commercial questions ("a good sunscreen for oily skin with delivery in Athens"), most assistants run a web search, read a handful of pages and synthesise an answer. If your pages are blocked or vague about what you sell, they will not be picked.
- Structured product data. Price, availability, brand, identifiers (GTIN/MPN), ratings. Machine-readable data is trusted more than prose.
- Third-party mentions. Comparison articles, forums, Reddit threads, reviews, buyer guides, even your Skroutz profile. Assistants effectively "vote" based on how often, and in what context, your name appears.
- Entity clarity. Is it obvious who you are, where you operate, what you sell, how shipping and returns work? If not, the model will not risk recommending you.
The good news is that almost all of this is also good classic SEO. The difference lies in the details and the format.
1. Let the right bots in
Check your robots.txt. Plenty of shops block AI crawlers without knowing it, usually because of a security plugin or a Cloudflare setting. The names worth knowing:
- OAI-SearchBot and ChatGPT-User: used for ChatGPT search and live page visits.
- GPTBot: OpenAI's crawler for model training data.
- Google-Extended: controls whether your content is used for Gemini. AI Overviews rely on the regular Googlebot.
- PerplexityBot, ClaudeBot, Bingbot.
It is a perfectly reasonable choice to block training crawlers and allow search crawlers. What you want to avoid is shutting everything out by accident. Check your firewall and CDN rules too, not just the file.
2. Product pages machines can read
An AI does not see your lovely photos. It sees text and data. On every product page:
- Complete Product schema in JSON-LD with price, currency, availability, brand, GTIN and, where you have them, aggregateRating.
- Core content (title, price, description, specs) present in the initial HTML, not injected later by JavaScript.
- A description that answers real questions: who it is for, how it is used, how it differs from the similar product next to it, dimensions, materials.
- A short FAQ per product or category, written in the words your customers use on the phone.
AI tools do passage retrieval: they pull a specific snippet, not the whole page. One clear paragraph answering one question is far more likely to be quoted than 800 words of generic copy.
3. Category pages as buyer guides
Many questions to assistants are comparative: "which robot vacuum for a home with pets?". A category page with two or three paragraphs of guidance (how to choose, what matters, who each subcategory suits) is exactly the kind of content that gets cited. Keep it short and useful, above or below the product grid.
4. Be clear about the shop itself
Put the following on stable, indexable pages:
- Who you are and how long you have been trading, with real company details.
- Shipping costs and times by region, which couriers you use, whether you accept cash on delivery, whether you deliver to parcel lockers.
- Your returns policy in plain language.
- Organization schema with name, logo, contact details and
sameAslinks to your official profiles.
When someone asks "which shop ships quickly to Crete?", that information has to be written down somewhere.
5. Build mentions off-site
This is where AI visibility differs most from traditional SEO. Backlinks still matter, but so does a plain mention of your brand in the right context. In practice:
- Look after your Skroutz and BestPrice profiles: reviews, accurate details, replies to feedback.
- Keep a Google Business Profile up to date, even if you are online-only.
- Work with blogs and magazines in your niche on genuine buyer guides, not paid link placements.
- Take part in communities (groups, forums) by actually helping people, not by spamming.
6. Feeds: the most underrated move
If you already run a clean Google Merchant Center feed, you are halfway there. Google's AI Overviews and AI Mode pull product data from the Shopping Graph, which Merchant Center feeds into. Other AI platforms are also moving towards accepting merchant product feeds, so a complete, accurate feed will serve you wherever the landscape settles.
How to tell if it is working
Measurement is still imperfect, but you can:
- Track referral traffic in analytics from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com.
- Write down 20–30 questions your customers would ask and run them monthly in three or four assistants, noting whether and how you appear.
- Listen for customers saying "I found you on ChatGPT" on the phone or in order notes.
We automate that last kind of monitoring with our Visibility app, because doing it by hand gets tedious fast.
Quick checklist
- robots.txt and firewall allow AI search crawlers
- Complete Product schema on every product
- Core content in the initial HTML
- Descriptions that answer real questions
- Buyer-guide copy on key categories
- Clear, current shipping, returns and About pages
- Organization schema with sameAs
- Clean Merchant Center feed
- Monthly check of your presence in ChatGPT, Gemini and Perplexity
The bottom line
AI search does not replace SEO. It demands a more careful version of it. The shops that get recommended are the ones with clean data, genuinely useful content and a real reputation beyond their own site. If you want to see where yours stands, have a look at our AI visibility service or drop us a line for a first review.