How to Rank on ChatGPT: GEO Playbook for UK Businesses 2026

Ranking on ChatGPT works nothing like ranking on Google - third-party mentions, answer-first formatting, and placement in listicles matter far more than keyword optimization.

That last point hits harder when you see it in practice. I asked ChatGPT this week how to rank on ChatGPT. The response cited zero websites. No links, no sources, nothing pulled from Google. Just a generic answer about "crafting clear prompts," written straight from training data. The sites that should have owned that answer weren't there - because they hadn't done the work to get mentioned in the places ChatGPT actually learns from.

Ranking on ChatGPT means becoming part of the consensus the model draws from, not winning a keyword position. You are not competing for a slot on a results page. You are trying to be the source ChatGPT reaches for when it constructs an answer, whether it searches the live web or not. The discipline that covers all of this is called Generative Engine Optimization (GEO) - and the playbook most UK agencies are selling you was written for a different game.

This is what actually moves the needle, based on what I have tested for Nest Content and what the largest public studies of AI citations have found. If you want the shortest possible version: get into the listicles your competitors are already in, rewrite your key pages in an answer-first format, and stop paying for things like llms.txt files that do nothing.

What ChatGPT actually cites (and what it does not)

ChatGPT answers in two different modes and most articles about ranking get this wrong.

Mode one: training data only. The model generates an answer from what it already knows, without reaching out to the web. This happens on conceptual questions, opinion questions, and most queries that do not clearly need fresh information. Ranking here means being mentioned often enough across the training corpus that the model has learned your brand, your positioning, and what you do. You get no click, no citation, but you do get brand mention.

Mode two: live web retrieval. The model triggers a web search, pulls in real pages, and cites them as sources in the answer. This is where you can earn a clickable link. It happens more often on "best of" queries, product comparisons, recent events, and "how much does X cost" style questions.

The split matters because tactics differ. For mode one you need widespread mentions on third-party sites. For mode two you need to rank for the related Google query, because ChatGPT now uses Google's index after switching from Bing in mid-2025.

Kevin Indig analysed 1.2 million ChatGPT responses and 18,012 verified citations. The biggest pattern: 44 percent of ChatGPT citations come from the first 30 percent of a page's content - what Indig calls the "ski ramp" effect, because citation rates drop sharply as you move down the page. Separate research across 2 million ChatGPT sessions found that 72 percent of cited blog posts included a concise 40 to 60 word "answer capsule" placed directly after a question-form heading. Ranking in Google opens the door to being cited. Answer-first formatting is what actually gets you through it.

Seven things that genuinely move the needle

These are in rough order of impact, based on citation studies and what I have tested directly.

1. Write answer-first, not intro-first

Your first 40 to 60 words under every heading need to contain the direct answer. The 44 percent "ski ramp" finding means if your opener is a throat-clearing intro that sets up the topic, you lose the citation before the model ever reaches your actual answer. Indig's research also found heavily cited text averaged 20.6 percent proper nouns against 5 to 8 percent in typical English text, so include specific tool names, company names, people, and places rather than generic descriptors.

Rewrite your existing articles like this: treat each H2 as a question, and make the first paragraph a self-contained, definitive answer in the "X is Y" format. Expand with context after. The old inverted pyramid from journalism works here, not the "build up to the reveal" style that became popular in SEO blogging.

2. Get into three to five listicles, not just rank number one

Peec AI analysed 232,000 AI citations across 13,000 listicles over 12 weeks on ChatGPT, Perplexity, Google AI Mode, Copilot, Gemini, and AI Overviews. Listicles accounted for 22 percent of all citations, articles for 17 percent, and product pages for 14 percent. Together those three formats cover more than half of everything AI engines cite. When ChatGPT answers "what is the best X for Y in the UK", it does not read one article and pick a winner. It reads four or five listicles and reports the names that appear in multiple.

This is why consensus beats ranking. A brand that appears in five mid-tier roundups gets cited more than a brand that ranks first in one flagship article. Power Digital mentions this in passing. Almost no UK agency explains how to actually get into those listicles. I've broken down when that placement work justifies paying for help in should you hire a GEO agency or do it yourself.

My approach: identify the five to ten listicles ChatGPT pulls from for your target queries, find the contact email for the editor or author, send a one-paragraph pitch with a unique angle you can add to their piece. Our automation-first approach to SEO, covered in AI SEO tools, is what we lead with for our own outreach. Response rate for these pitches tends to sit around 15 to 20 percent when the angle is genuinely useful to their readers.

3. Structure for extraction, not just reading

The pages that get cited look similar across industries:

• Question-form H2s that match how people phrase their prompts
• Short paragraphs, usually two to four sentences
• Bullet lists and tables where comparisons are involved
• At least one entity-dense paragraph with named brands, tools, people, and places
• Definitive language ("X is Y because Z") rather than hedged academic prose

The goal is self-contained paragraphs that still make sense when pulled out of context. If the first sentence of a paragraph starts with "this means that..." and references something earlier, an AI cannot extract it cleanly.

4. Build entity authority, not just backlinks

Emerging pattern from multiple citation studies: traditional backlink metrics are a much weaker signal for AI citations than they are for Google rankings. A page can have strong domain authority by classic SEO standards and still get ignored by ChatGPT if it has no entity signals.

What does correlate positively with AI citations: brand mentions across multiple trusted sites, even without links. Schema markup that names your organization, author, and credentials. Author bylines with verifiable credentials on platforms like LinkedIn. Consistent entity information across directories, your site, and your Google Business Profile.

This is why I recommend working on content differentiation and author positioning before doing any outreach. An anonymous "admin" byline on an otherwise decent article gets skipped for ChatGPT citations far more than a named expert with a verified track record.

5. Allow GPTBot, ChatGPT-User, and OAI-SearchBot in robots.txt

Blocking these crawlers is the single biggest own-goal I see. Some sites block them on principle because they read a piece about AI training data theft. Others block them accidentally because their agency set up aggressive bot filtering and never revisited it. Check your robots.txt file for these three entries right now:

User-agent: GPTBot
Disallow: /

User-agent: ChatGPT-User
Disallow: /

User-agent: OAI-SearchBot
Disallow: /

If any of them show Disallow: /, you have made yourself invisible to ChatGPT's web search. Remove those lines or change to Allow: / and confirm the change via a fresh fetch. Some platforms like Cloudflare add these blocks by default in their bot management settings; check there too.

6. Use Google, not Bing, for keyword research

A lot of older guides tell you to optimise for Bing because ChatGPT uses the Bing index. That changed in mid-2025. ChatGPT now uses Google's index for live search, which means your Google Search Console data is the right starting point for which queries you can reasonably win.

Open GSC, filter to queries where you already rank between positions 5 and 20, and identify which of those queries are likely to trigger AI answers. If you want a broader view of the SEO tool landscape for this kind of work, we've compared the options that actually surface AI visibility data. Commercial "best of" queries, "X vs Y" queries, and "how much does X cost" queries are the most likely triggers. These are the queries where optimising for AI visibility actually has a chance to compound on top of your existing rankings.

7. Earn brand mentions on LinkedIn, Medium, and industry forums

Unlinked brand mentions on trusted third-party sites show up consistently in what ChatGPT surfaces. The practical tactics that work for small businesses:

• LinkedIn articles under a named expert byline. Fastest-rising citation source in my own tracking.
• Forbes Council, Medium, or industry publications. Even a single syndicated article creates a second touchpoint.
• PRNewswire or RNS releases for anything genuinely newsworthy (funding, product launch, partnership).
• Industry directories and review sites like Clutch, Capterra, G2, TrustRadius - whichever matches your niche.

The pattern across all of these: get your brand name onto a site that a ChatGPT query is likely to surface, even if the mention is small. A one-sentence mention on a credible third-party domain is worth more to AI visibility than a dedicated 2,000-word page on your own blog.

What does not work, even if you keep seeing it recommended

Three tactics that still show up in most "rank on ChatGPT" articles and do not produce measurable results:

llms.txt files. A small but vocal group of SEO consultants is selling llms.txt implementation as a GEO service. No evidence that any major AI platform reads these files in a way that changes citation behaviour. Not OpenAI, not Anthropic, not Google. Skip it.

Buying backlinks. Counterproductive for AI citations specifically. Do not waste budget here if your goal is AI visibility. The same budget spent on listicle placement or LinkedIn authorship delivers 5 to 10x more AI surface area.

Keyword-stuffing with "ChatGPT" variations. Writing "ChatGPT SEO", "ChatGPT optimization", "rank on ChatGPT" 40 times across a page does not help. AI models interpret content semantically. What hurts you here is keyword stuffing making the content look promotional or low-trust to both Google and ChatGPT.

How to monitor your ChatGPT visibility (cheaply)

You cannot improve what you do not measure. The expensive way to monitor AI visibility is a dedicated tool at £100 to £400 per month. The cheap way uses the DataForSEO LLM Responses API at roughly $0.03 per query.

Here is the workflow I run for Nest Content and for clients:

1. Write ten prompts that match how real buyers would ask ChatGPT for a recommendation in your niche. Keep them natural, not keyword-stuffed.
2. Run each prompt twice per month via the API.
3. Parse the response for mentions of your domain and any cited URLs.
4. Log which competitor domains get cited and which listicles or directories show up as sources.
5. After eight weeks, compare your mention count and measure which tactics moved it.

Total cost: about £6 per month for 20 queries. No monthly tool subscription needed.

The most useful output is not your own mention count. It is the list of cited sources. Those listicles and directories are your real outreach targets, because they are where ChatGPT looks first for your category.

Track ChatGPT referral clicks in GA4 (zero cost)

The DataForSEO method above tells you whether ChatGPT mentions you. GA4 tells you whether those mentions drive actual clicks to your site. Both are worth running.

ChatGPT passes referral data when users click cited links. In GA4, set this up in three steps:

1. Go to Reports > Acquisition > Traffic acquisition
2. Add a filter: Session source matches regex chatgpt|openai
3. Filter by Session default channel group = Referral

You will see sessions arriving from chatgpt.com and openai.com. This shows which pages generate clicks, not just citations. A page that gets cited frequently but produces zero GA4 referral sessions has a mismatch between its snippet and what users expect - fix the page title or meta description first before touching the content.

What the realistic timeline looks like

How fast can you rank on ChatGPT? For live web retrieval, well-optimised content can appear in cited answers within 48 to 72 hours. For training data inclusion - the offline mode where ChatGPT answers from memory without running a web search - expect a full model update cycle, which runs quarterly at best. Most businesses see meaningful movement at the 4 to 8 week mark, provided they tackle listicle placement and robots.txt fixes together rather than in sequence.

Honest expectations, because this is where most agencies lose trust:

• Schema and robots.txt fixes: visible to ChatGPT's live search within 1 to 4 weeks.
• Answer-first rewrites of existing ranking pages: measurable mention changes within 4 to 8 weeks.
• Listicle placement from outreach: 30 to 60 days from the first pitch to publication. Some publications sit on drafts for months.
• Training data inclusion: quarterly at best. Brand-new domains with no third-party mentions will not appear in ChatGPT's "offline" training-only answers for at least a full model update cycle.
• Significant visibility lift: realistically 3 to 6 months of consistent work.

If someone is promising ChatGPT visibility in two weeks, they are either testing a single trivial query or lying to you.

What this looks like in practice: a Bristol cybersecurity consultancy

One example from our own client work. A three-person B2B cybersecurity consultancy in Bristol came to us with a specific problem: when their prospects asked ChatGPT for "best cybersecurity consultants for small UK manufacturers", they did not appear. Their two main competitors were cited in seven out of ten queries we tested.

We audited their presence across the listicles ChatGPT was pulling from. They were in zero of them. Their own site had a thorough services page but no bylined expert authorship, no schema, and a robots.txt blocking AI crawlers on a rule inherited from a previous agency.

Over six weeks we did four things. Fixed the robots.txt and added Organization and Person schema with verifiable founder credentials. Rewrote their three service pages in answer-first format. Pitched and placed the founder on two industry listicles and one Medium publication. Published one LinkedIn article from the founder explaining their manufacturing-specific methodology.

At the eight-week re-test, they appeared in four out of ten prompts. Not full parity with the incumbents yet, but a meaningful move from invisible to surfaced. The listicle placements were the single biggest driver. The robots.txt fix alone would have done nothing without the third-party mentions to back it up.

The honest read on Nest Content's own position

Worth saying out loud because it would be strange to write this article without it. Nest Content has had 4,011 ChatGPT referral sessions over the last 16 months. That sounds like a win until you look at where they landed.

74 percent of that traffic went to one free AI content detector tool. Another 9 percent to a trending topics tool. Only a handful of sessions hit a service page or a buyer-intent blog article. In GA4, the sessions landing on tool pages converted to zero new customers. The users arrived, used the free tool, and left.

When I ran the citation audit on buyer-intent queries specifically - "best managed SEO service UK", "UK small business SEO agency", "how much does managed SEO cost" - Nest Content appeared in zero out of ten. The ChatGPT traffic we had was real. It was also wrong-ICP traffic. Free-tool users, not people looking to hire an SEO agency.

That is why we killed the AI content detector tool on 12 April. Wrong-ICP traffic at scale is not a win. It pulls crawl budget and brand signal away from the pages that would bring qualified prospects, and it trains the model to associate our domain with "free utilities" rather than "managed SEO service". We are now running the same playbook I just described on our buyer pages: listicle placement in UK SEO agency roundups, LinkedIn authorship from me personally, answer-first rewrites of our service pages, and monthly citation monitoring for the ten queries that matter.

Two lessons from our own mess. Getting cited by ChatGPT is not the same as getting cited for the right queries - volume without intent is worse than silence. And free-tool pages that attract wrong-ICP AI traffic are worse than useless, because they dilute your brand signal in exactly the direction you do not want. If you are reading this inside a ChatGPT answer six months from now for something like "best managed SEO service UK", something worked.

If you want the same playbook run for your business instead of doing it yourself, our managed SEO service covers AI visibility alongside traditional SEO.

Frequently asked questions

How is ranking on ChatGPT different from Google SEO?

Google ranks pages by domain authority, backlinks, and keyword relevance. ChatGPT surfaces sources by consensus - it favours content that appears across multiple trusted sites, uses answer-first structure, and contains named entities. A page that sits on page 3 of Google can still get cited heavily in ChatGPT if it appears in the right listicles. The skill set overlaps but the levers are different.

Does ChatGPT use Bing or Google to search the web?

ChatGPT switched from Bing to Google's index for live web search in mid-2025. Use Google Search Console to identify ranking opportunities, not Bing Webmaster Tools. Both platforms still crawl your content independently, so staying indexed in Bing costs nothing and does no harm - but Google is where the citation opportunity lives.

Do I need schema markup to rank on ChatGPT?

Schema is not required but it helps. Organization, Person, and Article schema give ChatGPT clear signals about who created the content and what it covers. It is a one-time technical fix with compounding returns. Priority order: fix robots.txt blocks first, add schema second, then focus on listicle placement.

Will keyword stuffing help me appear in ChatGPT results?

No. ChatGPT interprets content semantically. Repeating "best SEO agency UK" forty times on a page does not increase your citation rate - it reduces it, because over-optimised content reads as low-trust to both Google's indexer and ChatGPT's retrieval layer. Write for the question behind the query, not the query itself.

What is the difference between GEO and SEO?

Generative Engine Optimization (GEO) is the practice of optimising content and brand presence so that AI engines like ChatGPT, Perplexity, and Google AI Mode cite you in generated answers. Traditional SEO targets a ranked list of links. GEO targets the single synthesised answer an AI produces - a narrower target, but one with outsized influence on the shortlists buyers form before they ever visit your site.

Start with the two cheapest moves

You do not need a 12-month roadmap to see if this works for your business. This week, do two things:

1. Open your robots.txt file and remove any blocks on GPTBot, ChatGPT-User, and OAI-SearchBot. A basic Chrome extension stack can help you verify the fix on your own pages without running curl.
2. Identify the three listicles that rank for your primary "best of" query in the UK. Find the editor's email. Send a short pitch with one unique angle you could add.

If nothing else moves in 60 days, the rest of the playbook probably will not help you either. If those two moves start surfacing you in AI answers, you have proof that the rest is worth investing in.

And if you would rather skip the DIY route and have the full playbook run for your business - listicle outreach, answer-first rewrites, schema and entity work, monthly citation monitoring - that is exactly what we do at Nest Content. Book a free 15-minute strategy call and I will pull up your domain, show you where your competitors are cited and you are not, and lay out what we would do in the first 30 days. No slides, no pitch deck, no commitment.