Search Intent: The Mistake That Makes Your Content Invisible to AI

TL;DR: AI engines do not cite all content, whatever its quality or the size of the brand. The first filter is search intent: commercial and transactional queries trigger citations, pure informational ones almost never do. Here is how to sort your queries before writing, proven across 5 AIs and 3 sectors, and why content that is never cited remains your best authority foundation.

You produce content, you optimize your site, and yet the AIs never cite you. The mistake that wipes out all these efforts: getting the search intent wrong. Many bet on content that AI engines will never cite, whatever its quality, whatever the size of the brand.

Because before your size, the first filter is the intent behind the query. On some questions, the AI cites sources in droves. On others, it cites nobody: not the giants, not you, not your competitors.

Here is how to recognize the queries that trigger a citation, proven by asking the same question to Google, ChatGPT, Claude, Gemini and Perplexity, across three sectors. And why content that is never cited can still remain your best investment.

Quick vocabulary. SEO aims to rank your pages well on Google. GEO (Generative Engine Optimization) does not play on the same field: to be cited by ChatGPT, Gemini or Perplexity, you must build your authority on your site and everywhere else: mentions, reviews, comparisons, communities. It is no longer a battle of pages, it is a battle of presence.

AI Citation Is Rare, and Concentrated on a Few Sources

First reality to internalize: getting cited by an AI is rare. Citations concentrate on a handful of sources. According to Ahrefs (analysis of 55.8 million Google AI Overviews), the top 50 domains capture 28.9% of all citations.

And it is almost always the same families of sources:

• encyclopedic sites (Wikipedia),
• community platforms (Reddit, Quora),
• YouTube,
• official and government sites,
• recognized media outlets.

Worse still: on a large share of the queries your customers type, the AI cites no brand at all. Not yours, not your competitors'. Knowing when that happens means no longer spending where the citation is impossible to win.

Citation therefore goes to sources judged reliable, not to everyone. Two pieces of good news for what follows: this reliability can be built, and above all it depends first on the intent of the query.

The 4 Search Intents, in Plain Terms

Intent is what the person really wants to get by typing their question. Marketing distinguishes four. Knowing them means knowing in advance whether an AI will need to cite you.

The rule: the more a query pushes to prove, compare, quantify or update, the more the AI cites. Commercial and transactional are the fields where citations are won. Pure informational, virtually no citation, unless it involves verification.

The AIs confirm it themselves. ChatGPT: "The more the query involves proving, justifying, comparing or updating, the higher the probability of relying on sources." Claude cites for "verifiable facts: numbers, statistics, dates", and becomes the source itself for advice.

Why Intent Does Not Mean the Same Thing in SEO and in AI

The concept of intent comes from SEO, but AIs use it differently. That is the nuance that changes your whole strategy.

In SEO: Intent Picks a Page

Google interprets intent to rank the right pages, and it always returns a list: something necessarily ranks. Intent tells you what type of page to produce in order to rank.

In AI: Intent Decides Whether to Cite

The AI does not match a keyword to a page. It reads the overall meaning of your request and asks itself a question Google does not: "do I need to prove my answer?"

• If yes, it goes looking for external sources (that is RAG, retrieval-augmented generation) and cites them.
• If no, it answers from its training knowledge, its internal memory, without retrieving anything, and therefore without citing anyone.

This need to verify depends on the subject and the sector. A recent fact, a price, or a regulated point triggers the search. A general definition does not.

The Three Faces of Informational

"Informational" covers three very different cases, and only one never cites:

• Pure informational (stable notion): "what is GEO?", "what are Ray-Ban Meta glasses for?". The knowledge does not move, the AI answers from memory: no source.
• News informational (the info evolves): "Meta AI's latest features", "what OpenAI announced this week". Its memory is dated: it cites.
• Verification informational (proving a claim): "is it true that...", "is it legal to film in the street?". Authority is needed: it cites (especially on sensitive or regulated subjects).

So it is not informational content that kills the citation, it is pure, stable informational content.

The shift to remember:

• In SEO, you optimize for a query.
• In GEO, you optimize for an intent, and for the precise moment when the AI decides it needs proof.

Tests run in June 2026. AI answers are not deterministic: they vary by day, user, and history. What matters are the underlying logics, not the exact positions.

My Tests: Intent Decides the Citation

I asked the same questions to the five AIs while keeping the same product (the Ray-Ban Meta glasses) and changing only the intent. The result is crystal clear.

• Transactional ("Ray-Ban Meta price"): the AIs cite heavily: product pages, marketplaces (retailers, price comparison sites), Reddit reviews, comparison engines. This is where citations are won.
• Simple informational ("what are Ray-Ban Meta glasses for?"): no AI cites a source, they answer from memory. Only Perplexity, an AI search engine, is the exception. No citation to win, only awareness.
• Verification informational ("is it legal to film with them?"): citations come back (data protection authorities, legal articles).
• Commercial ("best smart glasses 2026"): almost all cite, and it is the comparison guides that win the spot, not the product pages.

One rule above all: at equal intent, the more precise the query, the more the AI needs a concrete source. "Best hotel" stays vague; "best hotel in Bordeaux for a wedding anniversary" forces the AI to cite a specific place, and that is where a brand stands out.

Different Sector, Same Logic, Different Sources

I repeated the exercise on hospitality, then on SEO/GEO tools. The intent hierarchy holds everywhere. But the landscape of sources is completely redrawn from one sector to another.

• Sources vary completely by sector. Glasses: marketplaces and price comparison sites. Hospitality: almost everything goes through Booking. SEO tools: Semrush and review platforms. You must map the sources of your sector.
• A dominant player often sweeps the citation (Booking in hotels, Semrush in SEO). If you are absent from the sources it feeds, you do not appear.
• The verification trigger is not universal. It worked on the legal question about the Ray-Bans, but stayed silent in hospitality and SEO. Hypothesis: it only triggers citations on regulated or sensitive subjects (law, health, finance).

Three sectors, three source maps, one single intent logic.

A word on the impermanence of answers. Nothing is fixed. From one person to another, the same question does not give the same answer. Depending on your conversation history, your saved memory and your profile, the AI personalizes what it displays and what it cites. Two users can type the identical query and get different sources. That is why you must look at the underlying logics, not a screenshot at a given moment.

And this is where a smaller brand gets its chance. On broad queries ("best hotel in Bordeaux"), the authority aggregators (Booking, TripAdvisor) lock up the citation. But on ultra-precise sub-queries ("hotel in Bordeaux with parking and a seminar room for 12 people"), no giant is authoritative: the AI must find a specific source, and that source can be you.

Working on these sub-queries means bypassing the big players' lock.

And it fits perfectly, because that is exactly how AIs work. Google AI Mode uses a technique called query fan-out: instead of processing your query as is, the AI breaks it down into several sub-queries (often 8 to 12), retrieves sources for each, then assembles a single answer. It is therefore already hunting for precise sub-intents. Your job is to be the source that answers one of them.

Why AIs Cite Authority Sites, and How to Become One

In the tests, when the AIs mention a source, they cite authority sites first: data protection authorities, Booking, Semrush, major media. Hence the reflex "it's over for the small players". Except that authority, in an AI's eyes, is not a question of size but of presence.

Ahrefs studied 75,000 brands. The best predictor of visibility in AI answers is brand mentions across the web (0.664 correlation), nearly 3 times more than backlinks (0.218), and ahead of Domain Rating (0.326). YouTube mentions even come out on top (≈0.737) in the December 2025 follow-up. It is a correlation, not proof of causation, but the signal is clear.

Translation: the more your brand is talked about in the right places, the more the AIs treat it as a reliable source. Authority sites come first because they worked on their presence, not because they were born big.

The subtlety to grasp: working to become a source, to be mentioned and picked up elsewhere, is already working to be cited. Being sourced and being cited are the two ends of the same chain. Every mention earned today prepares a citation tomorrow.

If so many brands are not yet cited, it is not a fatality of size. It is that they do not speak up where it counts: Reddit, YouTube, customer reviews, trade press, comparison sites. The field is largely empty. Whoever settles in now takes a lead that the AIs record.

What Content to Produce, Where, for Which Intent

This is the question that wins or loses budget. Example: the YouTube video. YouTube is one of the sources most cited by AIs (Perplexity's favorite, ≈16% of its citations, Semrush 2025). A "Ray-Ban Meta review" video surfaces on a commercial intent, but never on "Ray-Ban Meta price": the AI wants a price in text, not an 8-minute video.

Reddit remains, across all platforms, the most cited source, ahead of YouTube, LinkedIn and Wikipedia (Semrush 2025; corroborated by Ahrefs on 78.6M searches).

"Extractable" means an AI reads and reuses your info effortlessly: a price written out in text, a clean table, structured data (markup that describes your price and product to the machine).

Which Intent to Bet On Depending on Your Goal

Not all intents serve the same purpose, and some bring no citation at all. Better to know before spending.

• Goal: leads and sales: bet on commercial and transactional. That is where AIs cite, and the citation brings you someone already in buying intent. Your weapons: sourced comparison guides and extractable product pages.
• Goal: awareness and expertise: that is informational. Be lucid, you will almost never be cited there, but this presence feeds your authority, and therefore your future citations on commercial queries.
• Goal: protecting your brand: work on navigational. Be the official, clean, structured source on your own name.
• Queries with no direct citation: general definitions, pure advice, validation outside regulated subjects. They feed your authority, not your immediate citation.

One constant, whatever the goal: work on your brand mentions. That is the fuel that moves you, over time, to the side of the sources AIs cite first.

Uncited Informational Content Is Not Useless: It Is Your Authority Foundation

A reading trap lurks here: "since informational content is not cited, might as well abandon it". Wrong, and this is the real paradigm to understand.

A pure informational article does not get cited, but it builds your authority, and it is this authority that gets your commercial content surfaced and cited. Every piece of substantive content feeds your brand mentions, the best predictor of citation. It does not get cited, it makes you citable elsewhere.

The real trap is not producing informational content. It is producing pure informational content, the kind that triggers no verification, then seeing that you are not cited and concluding that this type of content is useless. And worse still: convincing yourself that AI only cites the big players anyway, so there is no point in optimizing or running a presence strategy.

GEO Checklist

1. Sort your queries by intent before writing. Aim for commercial and transactional for citation, informational for authority.
2. Target specific queries, not generic ones.
3. Make your page extractable: price in text, sharp image, structured data.
4. Be present where the AI already searches: marketplaces, comparison sites, the aggregator that dominates your sector.
5. Take care of Reddit and review platforms, massively used by AIs for comparisons.
6. Create sourced comparison guides, the winning format on commercial queries in every sector tested.
7. Put the key info in the first lines: a large share of citations is extracted from the beginning of a piece of content.
8. Date and attribute your figures.
9. Work on your brand mentions: the best predictor of AI citation (Ahrefs, 75,000 brands), far more than backlinks.
10. Match the format to the intent: video and Reddit for commercial, structured page for transactional.

A Limitation to Own

Three sectors is not the whole world. The intent logic generalizes; it is a behavior of the models. What does not transfer goes beyond the mere names of the sources. Based on my tests, two things really change from one sector to another.

First, the typology of sources: the citation can be captured by fragmented marketplaces where you can slip in, by a single aggregator like Booking where it is nearly impossible, or by community and review platforms. Your room for maneuver changes completely.

Then, the citation behavior: a verification trigger effective on a regulated subject can stay totally silent elsewhere. On health, finance or law, expect more institutional citations; on B2B SaaS, expect Reddit and G2.

The 4-intent framework transfers. The source map and the triggers have to be redone for your sector.

To Remember: Which Mistake Makes Your Content Invisible to AI?

Direct answer: producing content without looking at search intent. The mistake that ruins all your efforts is betting on pure informational content (definitions, "how it works") where ChatGPT, Claude and Gemini answer from memory and cite nobody (only Perplexity, a search engine, is the exception). You can have the best page on the web: on those queries, you stay invisible. Informational content is not useless for all that; it builds your authority. The mistake is expecting the direct citation from it, or stopping there.

What really decides your citation:

• Intent. Commercial and transactional send the AIs looking for a source; pure informational does not, except for recent news or verification (a regulated point, a data protection authority).
• Specificity. On "best hotel in Bordeaux", Booking and TripAdvisor lock it up; on a precise sub-query, the spot opens up.
• Sector. Marketplaces (Amazon, price comparison sites) for a product, Booking in hospitality, Semrush and review platforms (G2) in SaaS. Reddit, YouTube and Wikipedia count everywhere.
• Presence. According to Ahrefs, brand mentions predict citation best, far ahead of backlinks. It can be built; an SMB can get there.

The shift: Google ranks you in a list of pages, the AI composes a single answer and decides whom to cite. Fix the intent mistake, position yourself where the AIs already search, and you move from invisible content to content that increases its chances of being cited.

Across all my tests, on five AIs and three sectors, a single variable decides the citation upstream: search intent.

Above all, remember this: working to become a source is already working to be cited.

At ForgR, the core of our tool maps the intents that trigger a citation, sector by sector.

FAQ: Getting Cited by AI Depending on Intent

Which search intents generate the most citations?

Commercial (comparing before buying), transactional (buying) and verification (proving a fact) intents generate the most citations. Pure informational, like a definition, generates virtually none: the AI then answers from memory, without a source.

What type of content has the best chance of being cited by AI?

Sourced comparison guides, reviews and extractable product pages, with the price and data in clear text. In my tests, on commercial queries, it is the "X vs Y" guides and rankings that surface first.

Is it true that good SEO is enough to get cited by AI?

No. A well-ranked, crawlable site remains a prerequisite, but AIs cite according to the intent of the query, not according to Google ranking. On many queries, no brand is cited at all, whatever its rankings.

Is it true that AIs only cite big brands?

No. According to Ahrefs (a study of 75,000 brands), brand mentions across the web predict citation best (0.664 correlation), far ahead of the number of backlinks (0.218). This presence can be built; an SMB can get there.

Where should you be present to increase your chances of being cited?

On Reddit, YouTube, Wikipedia, review platforms and the dominant aggregator of your sector. The top 50 domains alone capture 28.9% of citations (Ahrefs, 55.8 million AI Overviews).

About the Author

Paméla, Co-founder of ForgR

ForgR deploys around your site a network of niche topical blogs, optimized for SEO and GEO, that build your authority by mentioning you naturally, so that your brand becomes the source Google and the AIs cite. All of it, automated.