GEO is not the new SEO. It's what SEO was always supposed to be.

Every few years someone announces that SEO is dead. This time the claim has a better argument behind it. I have spent seven years in growth marketing, SEO and performance marketing, first inside B2B firms and now running Paradigm Media Networks, and I have never seen a shift reach clients' dashboards as quickly as AI answers have. When ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews answer the question directly, fewer people click through to a website.

That shift is real. The conclusion most people draw from it is wrong. GEO is not the new SEO. It is what SEO was always supposed to be: useful, specific, verifiable work that a machine can find and trust. AI search did not invent a new game. It removed the shortcuts from the old one, and the brands that were doing the real work all along are the ones getting quoted.

What actually changed when search started answering

Classic search showed you ten links and let you decide. AI search reads those links for you, decides what is true, and writes one answer. Your brand is either inside that answer or it is not,.

The behavioural data is already in. Pew Research Center analysed browsing data from 900 US adults in March 2025 and found that users clicked a traditional search result on 8% of visits when an AI summary appeared, against 15% when one did not. Clicks on links inside the summary itself happened on just 1% of visits. People were also more likely to end their browsing session entirely after a page with an AI summary.

So the unit of success moves:

The mistake to avoid here is treating the drop in clicks as a verdict on the channel. A buyer who reads an AI answer that recommends you, then searches your brand name two days later, shows up in your analytics as direct or branded traffic. The influence happened in the answer. Your attribution simply cannot see it.

The question is no longer "do we rank?" It is "when someone asks an AI about our category, does it know who we are, and does it trust us enough to say so?"

What didn't change: the work behind every citation

Here is my first contrarian point. The loudest GEO advice comes from people selling GEO, while the clearest statement on the subject comes from Google, which says plainly that there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisations necessary. A page needs to be indexed and eligible to be shown in Search with a snippet. That is the gate.

Look at what AI engines reward when they choose sources: clear answers placed early, original information such as first-hand process, data and opinion, consistent facts about who you are, and pages a crawler can fetch and parse without fighting JavaScript. That is a description of good SEO.

The academic evidence points the same way. In the paper that coined the term, Aggarwal and colleagues (published at KDD 2024) tested content changes against a generative engine and reported that GEO methods could boost visibility by up to 40%. The methods that worked were the ones a good editor would ask for anyway: adding statistics, adding quotations from credible sources, and citing sources. Keyword stuffing, the classic shortcut, performed worse than doing nothing.

A tactical example: take your best-performing service page and read only the first two sentences under each heading. If those sentences could have been written by any competitor in your category, an AI engine has no reason to choose yours. Rewrite them to include something only you know, such as a process step, a price range you publish, a constraint you have seen in real projects, or a named standard you work to.

Prompt mapping: the GEO strategy you can build in a week

Keywords were a proxy for intent. Prompts are closer to intent itself, because people type full questions into ChatGPT and Perplexity, add their context, then ask follow-ups. Google describes its AI features as using a "query fan-out" technique, issuing multiple related searches across subtopics to build one response. One prompt from your buyer becomes a cluster of retrieval queries, and you want to be the best answer for as many of them as possible.

Before you buy a GEO tool, do this manually for one week:

1. List the real questionsWrite down 20–30 questions a buyer asks before choosing a provider like you. Pull them from sales calls, enquiry forms, WhatsApp chats and proposal objections, not from a keyword tool.
2. Add the follow-upsFor each question, write the two follow-ups a sceptical buyer would ask next, such as cost, alternatives, risks or "is it worth it for a business my size?"
3. Run every prompt on every engineAsk each one in ChatGPT, Perplexity, Gemini, Copilot and Google with AI Overviews or AI Mode. Use a clean session and record the date.
4. Record who wins and whyNote which brands are named, which URLs are cited, and how you are described, if at all.
5. Turn gaps into a content and presence planEvery prompt where a competitor is cited and you are not becomes either a page to write, a page to restructure, or a third-party site where you need to appear.

At the end of that week you have a prompt map, a real competitor list (often different from the one in your pitch deck) and a list of source sites you need to be present on. That is a GEO strategy. Everything else is execution.

The common error at this stage is mapping only your own brand prompts. "Is [your company] good?" matters, but the prompts that decide shortlists are category prompts: "best SEO consultant for a B2B SaaS company in India", "how much does a Shopify store audit cost". If you only test your name, you only learn whether the AI knows you exist, not whether it would ever recommend you.

Make your pages easy to quote

Structure matters more than it used to, because engines lift passages rather than whole pages. A simple pattern works well: an H2 phrased the way people ask the question, then two or three sentences that answer it directly, then context, evidence, an example and the nuance. Microsoft's own publisher guidance for Copilot citations recommends clear headings, tables and FAQs, and supporting claims with evidence and data.

Entity consistency: make sure the machines agree on who you are

A language model assembles a picture from your website, your social profiles, business listings, directories, press mentions, reviews and forum threads. When those sources disagree about what you do, where you are, who founded you or what you are called, the model's confidence drops, and low-confidence entities get hedged or left out of answers.

In practice, the fix is unglamorous. Write one canonical description of your business: legal name, trading name, founder, location, services, the categories you want to be known for, and the year you started. Then make every public profile match it, word for word where the platform allows. On your site, express the same facts in structured data where it is honest, using Organization, Person, Article and FAQPage, with sameAs links pointing to the profiles you control.

Here is my second contrarian point: for many small and mid-sized businesses, entity work beats new content. Publishing your fortieth blog post does less for AI visibility than making sure your Google Business Profile, Bing Places listing, LinkedIn company page and top three industry directories all say the same thing. Microsoft explicitly lists maintaining accurate business information in Bing Places among its recommendations for appearing in Copilot answers. It is the cheapest GEO work there is, and the most often skipped.

A tactical example: search your brand name in Perplexity and ask "What does [brand] do and who runs it?" Compare the answer against your canonical description. Every factual error points to a source you need to correct or outweigh.

Technical access: let the right crawlers in

You cannot be cited by an engine that cannot read you. Each AI platform uses its own crawlers, and robots.txt rules written years ago for scrapers sometimes block the ones that now feed answers. The distinction that matters is between bots that collect training data and bots that power live search results.

The operational reality is that most technical GEO problems I find in audits are self-inflicted: a security plugin blocking unfamiliar user agents, a CDN rule that challenges every bot, or key content rendered only after JavaScript runs. Check your robots.txt and firewall logs against the crawler names above before you touch a single paragraph of copy; I cover the rendering side in technical SEO foundations.

Measuring share of answer

If the unit of success is the citation, you need a metric for it. I use share of answer: across a fixed set of prompts, on a fixed set of engines, how often is your brand named or your URL cited, compared with competitors? It is the GEO equivalent of share of voice, and like share of voice it is only useful when tracked over time with the same prompt set.

The tooling has started to catch up. In February 2026, Microsoft launched an AI Performance report in Bing Webmaster Tools, in public preview, showing total citations, average cited pages, the grounding queries the AI used to retrieve your content, and citation counts per URL across Copilot, Bing's AI summaries and select partners. If your site is not verified in Bing Webmaster Tools, do it this week. Google, by contrast, folds AI Overviews and AI Mode traffic into the overall Web search totals in Search Console's Performance report rather than breaking it out separately.

For ChatGPT, Perplexity and Gemini, a manual tracker still does the job. Each month, rerun your prompt map and log four things per prompt: whether you were mentioned, whether you were cited with a link, your position among the brands named, and the sentiment of the description.

The mistake to avoid is chasing daily fluctuations. AI answers vary between sessions and between days. Look at monthly trends across the full prompt set, and treat any single screenshot, good or bad, as an anecdote. This is a system, not a campaign, which is the argument I make more broadly in marketing systems over campaigns.

Where to start this month

Most businesses will either ignore AI search or panic about it. Both reactions leave space for the brands that treat it as what it is: a stricter, more honest version of the same game. You need a short, repeatable loop, not a new acronym.

In week one, build your prompt map and baseline your share of answer. In week two, fix crawler access and entity consistency, because those unlock everything else. In weeks three and four, restructure your five most important pages so each section opens with a direct, evidence-backed answer, and add the original data, process and opinion only you can provide. Then rerun the prompts, compare, and repeat monthly. If you want to speed up the research and writing parts of that loop, my notes on AI in marketing workflows show where AI helps and where it quietly makes content worse.

Build things worth citing. Make them easy to find. Measure where you show up. Repeat.

Frequently asked questions

What is the difference between GEO and SEO?

SEO aims to rank pages in search results and win the click. GEO, or generative engine optimisation, aims to have your brand named and your pages cited inside AI-generated answers from tools like ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews. The foundations overlap almost entirely: crawlable pages, helpful content, authority and clear structure. What changes is the unit of success, which moves from rankings and sessions to citations and share of answer, and the tolerance for thin content, which drops sharply.

Do I need special optimisation to appear in Google AI Overviews?

According to Google's own Search Central documentation, no. Google states there are no additional requirements and no special optimisations necessary to appear in AI Overviews or AI Mode. A page must be indexed and eligible to be shown in Google Search with a snippet. Beyond that, the usual guidance applies: helpful, people-first content, good page experience and technical accessibility. Be wary of anyone selling a secret AI Overviews technique that contradicts the platform's published position.

How do I get my website cited by ChatGPT?

Start by making sure OpenAI's OAI-SearchBot is not blocked in your robots.txt or firewall, because OpenAI says sites opted out of it will not appear in ChatGPT search answers. Then build pages that answer specific buyer questions directly in the first two or three sentences of each section, back claims with sources and original data, and keep your brand facts consistent across your site and third-party profiles. Finally, earn mentions on the sites ChatGPT already cites for your category, which your prompt map will reveal.

What is share of answer and how do I measure it?

Share of answer is the proportion of AI-generated answers, across a fixed set of prompts and engines, in which your brand is mentioned or your URL is cited, compared with competitors. To measure it, build a list of 20–30 real buyer prompts, run them monthly in each engine, and log mentions, cited links, position among named brands and sentiment. For Microsoft Copilot, Bing Webmaster Tools now offers an AI Performance report with citation counts and grounding queries, which supplements the manual tracker.

Does blocking GPTBot remove me from ChatGPT search?

Not according to OpenAI's crawler documentation. GPTBot is used for collecting content that may train OpenAI's foundation models, while OAI-SearchBot powers ChatGPT's search features. They are separate user agents with separate robots.txt rules. A business can disallow GPTBot to opt out of training and still allow OAI-SearchBot so its pages remain eligible to appear in ChatGPT search answers. Check both rules explicitly rather than relying on a blanket block of all AI bots.

Is SEO still worth investing in if AI answers reduce clicks?

Yes, because AI answers are built on top of search infrastructure. Google's AI features draw on its index, Copilot grounds on Bing, and ChatGPT and Perplexity run their own search crawlers. Pew Research found fewer clicks on results when AI summaries appear, so the value shifts from raw traffic towards being the source the answer trusts. The investment is still SEO, judged by a stricter standard and measured with citations and branded demand alongside sessions.

Sources

1. GEO: Generative Engine Optimization (Aggarwal et al.) — arXiv / KDD 2024, 2023–2024
2. AI features and your website — Google Search Central, 2026
3. Google users are less likely to click on links when an AI summary appears in the results — Pew Research Center, 2025
4. Introducing AI Performance in Bing Webmaster Tools Public Preview — Microsoft Bing Webmaster Blog, 2026
5. Overview of OpenAI crawlers — OpenAI, 2026
6. Perplexity crawlers — Perplexity, 2026