Sentinel InsightsGenerative engine optimization strategy:a plan that earns AI mentions

A generative engine optimization strategy is a plan for getting your brand named, described accurately and cited when people ask AI assistants about your category. A good one covers four things: the questions you want to show up for, the facts you publish about yourself, the independent sources that mention you, and a monthly check of what the assistants actually say.

It sits on top of SEO and PR, aimed at a new kind of reader. Search engines still index your pages, journalists still decide what gets covered, and AI assistants read the results of both before they answer.

What generative engine optimization is aiming at

Generative engines are tools that write an answer instead of returning a list of links: ChatGPT, Gemini, Claude, Perplexity and Google’s AI Overviews. When a buyer asks one of them “which payroll tools suit a 20-person agency”, the answer names a handful of brands and sometimes links to sources. Your goal is to be one of those brands, described correctly, with credible sources behind the mention.

If you want the definition in more depth, our explainer on what generative engine optimization is covers the term and where it came from. This post is about turning it into a plan.

What the research says works

The term was popularised by a 2023 research paper, “GEO: Generative Engine Optimization”, by researchers from Princeton University and IIT Delhi, presented at the ACM KDD conference in 2024. The authors tested different ways of rewriting web content and measured how visible each source became in AI-generated answers. They reported that the best methods could boost visibility by up to 40%.

The methods that helped most were adding quotations from relevant sources, adding statistics and citing sources. Keyword stuffing did worse than leaving the text alone. That fits what PR people have long known: specific, sourced, quotable material travels further than vague copy.

The paper studied page-level changes in a test setting, so treat the figures as direction rather than a promise. No study can tell you what a given assistant will say about your brand next month.

A generative engine optimization strategy in five parts

Here’s the plan we’d use for most brands. Each part feeds the next.

1. Choose the questions that matter

Write 30 to 50 questions your buyers ask before they buy. Group them by stage: problem questions, category questions (“best X for Y”) and comparison questions that name a competitor. These become the prompts you test every month.

2. Run a baseline

Run each prompt in three or four assistants. Record whether you’re named, your position, any factual errors and every source cited. Pay most attention to the sources. They tell you which publications, directories and review sites the assistants trust in your category.

3. Fix your own facts

Assistants read your site and profiles too. Make your About page, service pages and pricing plain and consistent, open each key page with a direct one or two sentence answer, and match your LinkedIn, directory and review profiles to the site. Add Organization, Service and FAQ schema where it fits.

4. Earn mentions in the sources assistants cite

This is the biggest lever and the slowest one. Use your baseline source list as a target list:

• Pitch data stories and expert commentary to the trade and business titles that keep appearing in answers.
• Contact the editors of the “best of” roundups the assistants cite, with a clear reason your product belongs.
• Book podcast appearances on shows your buyers follow, since transcripts and show notes are text that gets read.
• Encourage genuine reviews on the platforms that matter in your sector.

Sponsored features can reach an audience quickly, but they should always be labelled as paid. Our media placement service works that way.

A hypothetical example

Picture a hypothetical accounting software company for small US construction firms. Its baseline shows it’s named in a handful of answers, usually with the wrong pricing model, while two competitors appear in most of them. The cited sources are a construction trade title, two software review sites and a “best accounting tools for contractors” roundup on a business blog.

A sensible plan would fix the pricing wording on its site and review profiles, pitch the trade title a short survey of contractors’ cash-flow problems with the method published, offer the roundup editor a demo and a clear reason to include the product, and ask recent customers for honest reviews. None of that guarantees a mention, but each step adds a credible source an assistant could draw on.

5. Review monthly

Rerun the same prompts each month. Look for three changes: more mentions, more accurate descriptions and new sources citing you. Tie each change back to the work done, and shift effort toward what moves the answers.

Content that gives assistants something to quote

Following the research, write pages that contain the material assistants seem to favour. A simple checklist for each important page:

Never invent statistics or quotes to fill the table. A fabricated figure that gets repeated by an assistant becomes a reputation problem, and it’s the kind of error journalists notice.

How long a GEO strategy takes

Expect months. AI search tools that read the live web can reflect new coverage within weeks, while answers drawn from model training data change much more slowly. Most brands should plan for a quarter to set the baseline and fix their own facts, then two or three more quarters of steady coverage before patterns are clear.

The pressure to start is real. Gartner predicted in February 2024 that traditional search engine volume would drop 25% by 2026 as people turn to AI chatbots and other virtual agents. Brands that build a record of credible coverage now will have more for assistants to draw on later.

Budget for it the way you’d budget for PR. A steady flow of one or two good pieces of coverage a month, backed by pages that answer buyers’ questions plainly, tends to do more than a single burst of activity followed by silence. Assistants and the people who write the sources they read both reward consistency.

Common mistakes to avoid

• Chasing tricks. Hidden text, keyword stuffing and mass-produced pages don’t hold up, and the GEO paper found keyword stuffing did worse than doing nothing.
• Ignoring off-site sources. Your own site matters, but assistants lean heavily on what independent publications say about you.
• Changing the prompt list every month. If the questions change, you can’t tell whether your visibility did.
• Buying promises. Anyone guaranteeing AI mentions is selling something no one controls. We never guarantee them.

Where Sentinel fits

Many teams can handle the prompt testing and on-site fixes themselves. Earning steady coverage in the sources assistants cite is usually where they want support, and that’s the part digital PR was built for. Our AI Visibility Audit is a one-off $2,500 project, and the AI Visibility Programme runs the ongoing work from $4,500 a month. Details are on the pricing page.

To see how we plan and run the work, visit our AI search visibility page.