GEO: How to Get Cited by ChatGPT, Perplexity and AI Overviews

Your next customer may never look at your search ranking. They will ask ChatGPT, Perplexity or Google something like “which company should we hire to automate our support inbox?” and read a synthesized answer that names three or four providers. If you are one of them, you are on the shortlist. If you are not, you may not exist in that buying journey at all.

This shift is already measurable. According to 6sense’s 2025 B2B Buyer Experience Report, 94% of B2B buyers use large language models during their purchase process, most often to compare vendors, and they complete about 61% of their journey before they talk to a seller. At the same time, Google’s results page sends fewer clicks: Ahrefs found that when an AI Overview appears, the top organic result gets about 58% fewer clicks.

Generative engine optimization (GEO) is the discipline of earning a place inside those AI answers. This guide explains how assistants choose their sources, which levers actually move citations, what gets penalized, how to measure progress and what GEO costs in 2026.

What GEO is and how it differs from SEO

GEO is the practice of making your brand, pages and data the material that AI assistants retrieve, quote and recommend. You will also see the term AEO (answer engine optimization); for practical purposes it means the same thing.

GEO does not replace SEO. Assistants that search the web still pull pages from search indexes, so crawlability, speed and technical health remain the foundation. What changes is the scoreboard and the weight of each lever.

Why it matters now: the numbers

The click economy of search is shrinking, and several independent datasets point in the same direction:

• Ahrefs (published February 2026, December 2025 data, 300,000 keywords): when an AI Overview is present, the click-through rate of position one is 58% lower. In April 2025 the gap was 34.5% (Ahrefs, MediaNama summary).
• Pew Research Center (July 2025): users clicked a regular result in 8% of visits when an AI summary was shown and in 15% without one. Only 1% clicked a link inside the AI summary itself (summary via Omnibound).
• Seer: CTR on queries with AI Overviews fell to 1.3% in December 2025 and partially recovered to 2.4% by February 2026 (via Cognizo).
• The exception is branded search. Amsive data shows click-through on branded queries rose by 18.7% (Neil Patel).

Put together: generic informational traffic is being absorbed by AI answers, while people who already know your name still click. GEO works on both ends. It gets your brand named in the answer, and that naming builds the brand demand that still converts.

The buying side confirms it. 6sense reports that 94% of buying groups rank their shortlist before speaking to any vendor, and the vendor that is the favorite before first contact wins about 80% of deals. If an assistant helps build that shortlist, it is effectively part of your sales team.

How AI assistants choose what to cite

When an assistant answers with web search, it rewrites the question into several queries, retrieves candidate pages, extracts the passages that best answer each part and composes a response with citations. Three things follow from that process.

Ranking is not the same as being cited. Vendor data compiled by Vizup suggests that only about 12% of URLs cited by ChatGPT, Perplexity and Copilot rank in Google’s top 10. Assistants often cite a page that ranks fifteenth but answers the exact sub-question cleanly.

Format matters. The same research found that list-style articles (“best X for Y”) earn a larger share of citations for commercial queries, and that schema markup on its own does not raise citation rates.

Consensus matters. An assistant is more confident naming a brand that appears consistently across independent sources: review platforms, industry media, comparison articles, community discussions. One self-praising homepage is weak evidence.

The GEO playbook in seven steps

1. Build a prompt panel and measure your baseline

Before changing anything, find out where you stand. Write 50–100 prompts that real buyers would type, grouped by intent:

• Category: “best AI automation agencies for mid-sized companies in Europe”
• Problem: “how can a dental clinic stop missing phone calls”
• Comparison: “Intercom Fin vs a custom support agent”
• Price: “how much does an AI receptionist cost per month”

Run the panel in four to six assistants (for example ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews or AI Mode). For each answer, record whether you are mentioned, cited with a link or explicitly recommended, which competitors appear, and which sources the assistant cites. Run each prompt more than once: answers are non-deterministic, so a single run tells you little.

You can do this manually in a spreadsheet or use a tracker. Published prices in 2026: Otterly costs $29, $189 and $489 per month for 15, 100 and 400 prompts; Peec AI costs €85, €205 and €425; Profound starts at $99–499 with an enterprise focus (Canlah, CiteMe). Investors take the category seriously: Profound raised a $96 million Series C at a $1 billion valuation in February 2026, and Peec AI raised $29 million.

2. Make sure AI crawlers can read you

• Check robots.txt and your CDN or firewall. Some security setups block AI crawlers by default. Decide deliberately whether crawlers such as GPTBot and OAI-SearchBot (OpenAI), PerplexityBot, ClaudeBot and Google-Extended may access your public pages.
• Serve real HTML. Key content hidden behind client-side rendering, tabs, logins or PDF-only downloads is harder to retrieve and quote.
• Publish an llms.txt file that summarizes who you are and links to your most important pages. Among the 15 agency websites in our market review, Distyl already does this. It is an emerging convention, not a confirmed ranking signal, so treat it as cheap insurance.
• Keep facts consistent. Company name, what you do, for whom and where should read the same on your site, your profiles and directories.

3. Write answer-first, quotable passages

Assistants quote passages, not pages. Structure important pages so that every key question gets:

1. a heading phrased the way people ask it;
2. a direct answer in two to four sentences that makes sense on its own;
3. the detail, numbers and caveats underneath.

Compare two openings. Weak: “Pricing depends on many factors and every project is unique.” Quotable: “An AI receptionist for a single clinic location typically costs a setup fee plus a monthly fee that includes a block of minutes; self-serve tools start at about $49 per month, while integrated setups cost more.” The second gives the assistant something specific to cite.

Add FAQ sections built from questions your sales team actually hears. Put a date on data (“as of October 2026”) and link to the original source. Specific, dated and sourced statements are easier for an assistant to trust than adjectives.

4. Publish honest comparison and “best for” pages

Since list and comparison formats earn a large share of commercial citations, publish them yourself, and make them credible:

• “X vs Y” pages that state clear criteria and admit where the alternative is better;
• “best tools for [use case]” pages that include competitors and explain who each option suits;
• “alternatives to [platform]” pages for buyers who are leaving a tool;
• transparent “from” prices. In our review of 15 AI agency websites, 13 publish no prices at all. An assistant cannot quote a price you hide.

5. Earn mentions where assistants look

Brand mentions and digital PR are among the strongest GEO levers. Practical sources:

• industry publications, podcasts and newsletters that cover your niche;
• review and directory platforms relevant to your market, such as Clutch for agencies;
• partner and integration directories (CRM, helpdesk, automation platforms);
• guest articles and expert quotes with original data;
• community discussions where your team answers questions genuinely, without spam.

Each mention should describe you the same way. Consistency helps an assistant connect “the company that does X for Y” with your name.

6. Publish fresh first-party data

Assistants favor specific, current facts. Original numbers are the hardest content to copy: benchmarks from your own operations, anonymized project data, small surveys of your customers, regularly updated price overviews.

If you want to scale pages, use programmatic SEO only on top of your own data, with each page answering a distinct query. Wise, Zapier, Canva and Tripadvisor are the textbook examples. Roll pages out in batches of 50–200 and monitor indexing and AI visibility before publishing more (MADX).

7. Track monthly and iterate

Re-run your prompt panel every month. Useful KPIs:

• share of answers: in what percentage of panel prompts your brand appears;
• citations: how often your own URLs are linked;
• source map: which third-party pages assistants cite in your category (these are your PR targets);
• branded search volume over time;
• self-reported attribution: add “ChatGPT or another AI assistant” as an option to “how did you hear about us?”.

Be realistic about attribution. Many analytics setups cannot separate clicks that come from AI answers, so do not judge GEO by last-click traffic alone.

What gets penalized

GEO does not mean flooding the web with AI-written pages. Google has been explicit, as PPC Land documents:

• the scaled content abuse policy has applied since March 2024. It penalizes volume without value, regardless of whether people or AI produced it;
• in March 2024, 837 of 49,345 tracked sites (1.7%) were deindexed, together losing 20.7 million monthly visits;
• since May 2026, spam policies officially apply to AI Overviews and AI Mode;
• the September 2026 spam update started on 24 September and was still rolling out on 2 October;
• since December 2024, Google Ads rejects ads that lead to sites removed from search by manual actions. A content shortcut can cost you paid traffic too.

Template pages that merely swap keywords get deindexed. Pages that answer a distinct question with real data survive.

What GEO costs in 2026

Price is not the main risk. The main risk is paying for relabeled SEO: some agencies simply renamed their SEO service as GEO. A real GEO engagement starts with a measured baseline and reports the same panel every month.

A paid complement: ads inside ChatGPT

Organic citations are not the only way into AI answers anymore. ChatGPT ads launched on 9 February 2026 with a $60 CPM and a $200,000 minimum. On 5 May 2026 a self-serve ads manager opened to all US businesses without a minimum budget, with CPMs of roughly $25–60 depending on category and conversion bidding available since June (ALM Corp, Flyweel). Few agencies have real experience with the channel yet. Treat it as a test budget next to GEO, not instead of it: a paid placement disappears when the budget stops, while earned citations keep working.

Common mistakes

1. Buying “GEO” without a baseline. Without a measured starting point, you cannot tell progress from noise.
2. Believing guarantees. No one controls what an assistant says. A promise of “guaranteed placement in ChatGPT” is a red flag.
3. Mass-producing AI articles. It triggers spam policies and dilutes your expertise signal, the opposite of what assistants reward.
4. Treating schema or llms.txt as magic. Both are hygiene. Neither substitutes for being mentioned and being specific.
5. Hiding the specifics. No prices, no process, no numbers: nothing to quote.
6. Ignoring off-site reputation. If the only page praising you is your own, assistants have little reason to trust it.
7. Measuring only clicks. Visibility inside answers influences shortlists even when nobody clicks.

Where GEO fits in your growth system

GEO works best as part of a content system, not as a separate project. One expert recording can feed blog articles, FAQs and social posts at once; we describe that workflow in the content factory guide. The leads that AI visibility brings in still need a fast response, which is where AI chat funnels come in. To judge whether any of this pays off, put it into an ROI model for AI initiatives before you commit a retainer.

Conclusion

Search is splitting into two layers: a shrinking layer of clicks and a growing layer of AI answers that shape shortlists before any sales conversation. GEO is how you show up in the second layer. Measure a baseline across a fixed prompt panel, make your site readable for AI crawlers, write passages worth quoting, publish honest comparisons and your own data, and earn mentions where assistants look. Skip the shortcuts: guaranteed placements and mass-produced pages cost more than they return.

If you want to know where your brand stands today, our growth team can run a baseline across the main assistants and show you which sources they trust in your category. The technical side, from rendering to llms.txt, is handled by our build team. Book a free 15-minute audit and we will tell you honestly whether GEO is worth it for your business right now.