GEO for New Sites: How to Get Into AI Answers When Nobody Knows You Yet

Ask ChatGPT for a recommendation and it hands you a shortlist. Ask Google and an AI Overview answers before you scroll to a single blue link. Increasingly, being cited in those answers is how customers find you and it looks like a game only big, established brands can win.

It isn’t. We took Pimlico Bookbinding – a brand-new site with essentially zero authority and got it cited by AI, almost from scratch. This is how AI decides what to quote, the levers that move the needle fastest, how to measure whether any of it is working, and exactly what we did for Pimlico.

How AI Decides What to Cite

Here’s a simple model of what’s actually happening. When an AI assistant answers a question, it assembles the response from sources it can find, parse, and trust. Three things matter far more than raw domain authority:

• Clarity. Content that answers the question directly and unambiguously is easy for a model to lift. Vague, rambling pages get skipped.
• Corroboration. AI trusts what’s confirmed in more than one place. A fact that shows up on your site and a forum and a directory and a mention on a reputable industry site reads as true. One page claiming it alone doesn’t.
• Trust signals (E-E-A-T). Real authorship, real expertise, real experience. The models and the ranking systems feeding them – increasingly weight who is behind the content.

The liberating part: none of that requires a huge, aged domain. A new site that is clear, corroborated, and credible can get cited over a bigger competitor that is none of those things. (Reddit is a good example of a source AI leans on heavily – more on how we used it below.)

How to Build AI Visibility

The practical levers that move a no-name brand into AI answers:

• A real, active brand presence. Genuine social profiles and a consistent brand name and description everywhere. AI builds an “entity” for your brand by connecting mentions across the web – so give it dots to connect.
• An expert blog with genuinely useful content. This is the engine. Publish content that directly answers the questions your customers actually ask.
• Citation-friendly page structure. A short, direct answer at the top of every article, clear headings and an FAQ block that maps to real follow-up questions.
• Schema.org markup. Machine-readable structure: Article, FAQPage, author, Organization/Person – so AI can parse who wrote what and what each page is about.
• External mentions. PR, brand mentions, backlinks, interviews, and presence on the sources AI actually reads – niche industry sites, forums, and Reddit. (We wrote up how we earn links from Reddit without getting banned in a separate post)
• Consistency. Same brand name, same details, same entity everywhere. Fragmented mentions confuse the model, consistent ones compound.

Our Case: Pimlico Bookbinding, From Zero

Pimlico was the hard version of this problem – a brand-new site, no authority, no existing AI presence at all. Here’s exactly what we did.

1. Citation-friendly articles. Every blog post opened with a short, direct answer to the question in its title – the exact snippet an AI can lift. Then a clean, well-structured body, and an FAQ block (with markup) answering the follow-up questions people really ask.

2. Real E-E-A-T, not the performative kind. Every article was genuine expert content – real bookbinding, real first-hand experience, real photos of the work. Each post carried a named author, and the author’s profile linked out to their real social accounts. We weren’t signalling expertise; we were showing it.

3. Schema.org markup on everything. FAQPage and Article markup on every post, with the author attribute filled in, so the machines could see who wrote each piece and what it covered.

4. External optimization. We placed articles and interviews on relevant, thematic sites – each with a backlink or a brand mention and earned mentions and links from Reddit. This is the corroboration lever in action: the same brand, confirmed across the places AI reads.

The result: brand visibility climbed and AI citations grew from essentially zero. A new bookbinder started turning up in answers where, months earlier, it didn’t exist as far as any model was concerned.

How to Measure AI Visibility (2026 Tools)

You can’t improve what you can’t see and unlike Google rankings, AI citations don’t show up in a familiar dashboard. A whole category of AI visibility trackers has grown up to fill the gap. They work by firing large volumes of prompts at the AI engines and recording when and how your brand gets mentioned. The main players as of 2026:

• Semrush AIO – tracks ChatGPT, Perplexity, Claude, Gemini and DeepSeek against a database of 100M+ monthly LLM prompts.
• Ahrefs Brand Radar – monitors six AI engines, with custom AI-prompt tracking added in early 2026.
• Profound – enterprise-grade; fires millions of prompts daily across ~10 LLM engines, with a shopping/retailer module.
• Otterly.ai – budget-friendly brand-citation tracking across ChatGPT, Perplexity, Google AI Overviews, AI Mode, Gemini and Copilot.
• Peec AI – measures your “share of voice” across ChatGPT, Gemini and Copilot.
• Promptwatch – prompt-first: tracks the actual user questions across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek.
• SE Ranking (Visible) – brand scoring and competitor benchmarking across ChatGPT, Gemini, Google AI Mode and Perplexity.
• Evertune – sends 1M+ custom prompts per brand monthly to probe what the models actually “know” about you.
• Authoritas / KIME – customizable multi-engine tracking, with KIME adding an “action centre” that turns findings into recommendations.

You don’t need all of them. Pick one, set a baseline for the questions that matter in your niche, and watch the trend as you publish. The absolute number matters less than the direction of travel.

Your Playbook

If you’re starting from zero, do it in this order:

1. Lock down a consistent brand presence – name, description, real social profiles.
2. Start an expert blog. Answer real questions, lead each post with the direct answer, add an FAQ block.
3. Add Schema.org markup – Article, FAQPage, author, Organization.
4. Make the content genuinely expert – named authors, first-hand experience, real proof.
5. Earn external corroboration – niche sites, interviews, Reddit, brand mentions.
6. Pick one AI-visibility tracker, set a baseline, and measure the trend.

Do it in that order and the citations follow. Skip the trust and corroboration steps and you’ll stay invisible no matter how much you publish.

AI visibility isn’t reserved for household names. The models reward content that’s clear, credible, and corroborated – and a small, new site can do all three faster than a big, complacent competitor.