This publish was sponsored by Peec.ai. The opinions expressed on this article are the sponsor’s personal.
Step one of any good GEO marketing campaign is creating one thing that LLM-driven reply machines really wish to hyperlink out to or reference.
GEO Technique Parts
Consider experiences you wouldn’t moderately anticipate finding immediately in ChatGPT or comparable programs:
- Partaking content material like a 3D tour of the Louvre or a digital actuality live performance.
- Reside knowledge like costs, flight delays, accessible resort rooms, and many others. Whereas LLMs can combine this knowledge through APIs, I see the chance to seize a few of this site visitors in the interim.
- Matters that require EEAT (expertise, experience, authoritativeness, trustworthiness).
LLMs can not have first-hand expertise. However customers need it. LLMs are incentivized to reference sources that present first-hand expertise. That’s simply one of many issues to bear in mind, however what else?
We have to differentiate between two approaches: influencing foundational fashions versus influencing LLM solutions via grounding. The primary is basically out of attain for many creators, whereas the second presents actual alternatives.
Influencing Foundational Fashions
Foundational fashions are skilled on fastened datasets and may’t be taught new data after coaching. For present fashions like GPT-4, it’s too late – they’ve already been skilled.
However this issues for the long run: think about a wise fridge caught with o4-mini from 2025 which may – hypothetically – favor Coke over Pepsi. That bias may affect buying choices for years!
Optimizing For RAG/Grounding
When LLMs can’t reply from their coaching knowledge alone, they use retrieval augmented technology (RAG) – pulling in present data to assist generate solutions. AI Overviews and ChatGPT’s net search work this manner.
As search engine optimisation professionals, we would like three issues:
- Our content material will get chosen as a supply.
- Our content material will get quoted most inside these sources.
- Different chosen sources assist our desired end result.
Concrete Steps To Succeed With GEO
Don’t fear, it doesn’t take rocket science to optimize your content material and model mentions for LLMs. Really, loads of conventional search engine optimisation strategies nonetheless apply, with just a few new search engine optimisation ways you possibly can incorporate into your workflow.
Step 1: Be Crawlable
Sounds easy however it’s really an vital first step. When you goal for most visibility in LLMs, you’ll want to enable them to crawl your web site. There are lots of completely different LLM crawlers from OpenAI, Anthropic & Co.
A few of them behave so badly that they will set off scraping and DDoS preventions. In case you are routinely blocking aggressive bots, examine in along with your IT crew and discover a strategy to not block LLMs you care about.
When you use a CDN, like Fastly or Cloudflare, make certain LLM crawlers usually are not blocked by default settings.
Step 2: Proceed Gaining Conventional Rankings
Crucial GEO tactic is so simple as it sounds. Do conventional search engine optimisation. Rank nicely in Google (for Gemini and AI Overviews), Bing (for ChatGPT and Copilot), Courageous (for Claude), and Baidu (for DeepSeek).
Step 3: Goal the Question Fanout
The present technology of LLMs really does a bit of greater than easy RAG. They generate a number of queries. That is known as question fanout.
For instance, once I just lately requested ChatGPT “What’s the newest Google patent mentioned by SEOs?”, it carried out two net searches for “newest Google patent mentioned by SEOs patent 2025 search engine optimisation discussion board” and “newest Google patent SEOs 2025 mentioned”.
Recommendation: Test the standard question fanouts in your prompts and attempt to rank for these key phrases as nicely.
Typical fanout-patterns I see in ChatGPT are appending the time period “boards” once I ask what individuals are discussing and appending “interview” once I ask questions associated to an individual. The present yr (2025) is commonly added as nicely.
Beware: fanout patterns differ between LLMs and may change over time. Patterns we see in the present day might not be related anymore in 12 months.
Step 4: Preserve Consistency Throughout Your Model Mentions
That is one thing easy everybody ought to do – each as an individual and an enterprise. Ensure you are constantly described on-line. On X, LinkedIn, your personal web site, Crunchbase, Github – at all times describe your self the identical method.
In case your X and LinkedIn profiles say you’re a “GEO advisor for small companies”, don’t change it to “AIO knowledgeable” on Github and “LLMO Freelancer” in your press releases.
I’ve seen folks obtain constructive outcomes inside just a few days on ChatGPT and Google AI Overviews by merely having a constant self description throughout the online. This additionally applies to PR protection – the extra and higher protection you possibly can get hold of in your model, the extra seemingly LLMs are to parrot it again to customers.
Step 5: Keep away from JavaScript
As an search engine optimisation, I at all times ask for as little JavaScript utilization as potential. As a GEO, I demand it!
Most LLM crawlers can not render JavaScript. In case your principal content material is hidden behind JavaScript, you might be out.
Step 6: Embrace Social Media & UGC
Unsurprisingly, LLMs appear to depend on reddit and Wikipedia lots. Each platforms supply user-generated-content on nearly each subject. And because of a number of layers of community-driven moderation, a number of junk and spam is already filtered out.
Whereas each will be gamed, the typical reliability of their content material continues to be much better than on the web as an entire. Each are additionally commonly up to date.
reddit additionally offers LLM labs with knowledge into how folks talk about subjects on-line, what language they use to explain completely different ideas, and information on obscure area of interest subjects.
We are able to moderately assume that moderated UGC discovered on platforms like reddit, Wikipedia, Quora, and Stackoverflow will keep related for LLMs.
I don’t advocate spamming these platforms. Nevertheless, for those who can affect the way you and rivals present up there, you may wish to accomplish that.
Step 7: Create For Machine-Readability & Quotability
Write content material that LLMs perceive and wish to cite. Nobody has figured this one out completely but, however right here’s what appears to work:
- Use declarative and factual language. As a substitute of writing “We’re kinda positive this shoe is nice for our prospects”, write “96% of consumers have self-reported to be pleased with this shoe.”
- Add schema. It has been debated many instances. Not too long ago, Fabrice Canel (Principal Product Supervisor at Bing) confirmed that schema markup helps LLMs to know your content material.
- If you wish to be quoted in an already current AI Overview, have content material with comparable size to what’s already there. Whereas you shouldn’t simply copy the present AI Overview, having excessive cosine equally helps. And for the nerds: sure, given normalization, you possibly can after all use the dot product as a substitute of cosine similarity.
- When you use technical phrases in your content material, clarify them. Ideally in a easy sentence.
- Add summaries of lengthy textual content paragraphs, lists of evaluations, tables, movies, and different varieties of difficult-to-cite content material codecs.
Step 8: Optimize your Content material
The unique GEO paperIf we have a look at GEO: Generative Engine Optimization (arXiv:2311.09735) , What Proof Do Language Fashions Discover Convincing? (arXiv:2402.11782v1), and comparable scientific research, the reply is obvious. It relies upon!
To be cited for some subjects in some LLMs, it helps to:
- Add distinctive phrases.
- Have professional/cons.
- Collect person evaluations.
- Quote consultants.
- Embody quantitative knowledge and title your sources.
- Use simple to know language.
- Write with constructive sentiment.
- Add product textual content with low perplexity (predictable and well-structured).
- Embody extra lists (like this one!).
Nevertheless, for different combos of subjects and LLMs, these measures will be counterproductive.
Till broadly accepted finest practices evolve, the one recommendation I can provide is do what is nice for customers and run experiments.
Step 9: Keep on with the Information
For over a decade, algorithms have extracted information from textual content as triples like (Topic, Predicate, Object) — e.g., (Woman Liberty, Location, New York). A textual content that contradicts identified details could seem untrustworthy. A textual content that aligns with consensus however provides distinctive details is good for LLMs and information graphs.
So persist with the established details. And add distinctive data.
Step 10: Put money into Digital PR
Every part mentioned right here is not only true in your personal web site. It is usually true for content material on different web sites. One of the best ways to affect it? Digital PR!
The extra and higher protection you possibly can get hold of in your model, the extra seemingly LLMs are to parrot it again to customers.
I’ve even seen instances the place advertorials had been used as sources!
Concrete GEO Workflows To Strive
Earlier than I joined Peec AI, I used to be a buyer. Right here is how I used the instrument – and the way I counsel our prospects to make use of it.
Be taught Who Your Opponents Are
Similar to with conventional search engine optimisation, utilizing an excellent GEO instrument will typically reveal surprising rivals. Recurrently have a look at a listing of routinely recognized rivals. For many who shock you, examine by which prompts they’re talked about. Then examine the sources that led to their inclusion. Are you represented correctly in these sources? If not, act!
Is a competitor referenced due to their PeerSpot profile however you have got zero evaluations there? Ask prospects for a overview.
Was your competitor’s CEO interviewed by a Youtuber? Attempt to get on that present as nicely. Or publish your personal movies focusing on comparable key phrases.
Is your competitor commonly featured on high 10 lists the place you by no means make it to the highest 5? Provide the writer who created the checklist an affiliate deal they can’t decline. With the subsequent content material replace, you’re virtually assured to be the brand new primary.
Perceive the Sources
When performing search grounding, LLMs depend on sources.
Have a look at the highest sources for a big set of related prompts. Ignore your personal web site and your rivals for a second. You may discover a few of these:
- A group like Reddit or X. Turn out to be a part of the group and be part of the dialogue. X is your finest wager to affect outcomes on Grok.
- An influencer-driven web site like YouTube or TikTok. Rent influencers to create movies. Make sure that to instruct them to focus on the appropriate key phrases.
- An affiliate writer. Purchase your strategy to the highest with increased commissions.
- A information and media writer. Purchase an advertorial and/or goal them along with your PR efforts. In sure instances, you may wish to contact their industrial content material division.
You too can try this in-depth information on easy methods to cope with completely different sorts of supply domains.
Goal Question Fanout
After you have noticed which searches are triggered by question fanout in your most related prompts, create content material to focus on them.
By yourself web site. With posts on Medium and LinkedIn. With press releases. Or just by paying for article placements. If it ranks nicely in search engines like google and yahoo, it has an opportunity to be cited by LLM-based reply engines.
Place Your self for AI-Discoverability
Generative Engine Optimization is not non-compulsory – it’s the brand new frontline of natural progress. At Peec AI, we’re constructing the instruments to trace, affect, and win on this new ecosystem.
Generative Engine Optimization is not non-compulsory – it’s the brand new frontline of natural progress. We at present see purchasers rising their LLM site visitors by 100% each 2 to three months. Generally with as much as 20x the dialog charge of typical search engine optimisation site visitors!
Whether or not you’re shaping AI solutions, monitoring model mentions, or pushing for supply visibility, now’s the time to behave. The LLMs customers will belief tomorrow are being skilled in the present day.
Picture Credit
Featured Picture: Picture by Peec.ai Used with permission.
