When your brand never surfaces in ChatGPT's answers to "what's the best tool for X", the instinct is to write more content and hope. The faster path is to work in order: confirm the gap, then check the reasons AI skips a brand from cheapest to most involved. AI assistants read the HTML your server sends, prefer content they can extract cleanly, and lean on third-party sources to decide who belongs in a category - so the fix is rarely "more words" and usually "the right first step". CLEO is one option here: a Presence Engine that measures where you stand across AI answers and works to close the gap.
This article gives a starting order for a brand that is absent from AI recommendations and does not know where to begin.
My brand never appears when people ask ChatGPT for recommendations in our category and I don't know where to start fixing it - where do I begin?
Begin by making the invisible measurable, then fix in priority order:
- Confirm the gap. Run your category's real recommendation prompts through ChatGPT, Perplexity, and Google AI Overviews and record who gets named. You cannot fix what you have not measured.
- Check machine readability. AI crawlers read server-sent HTML and do not run JavaScript. A client-rendered site can be blank to them. View the raw source, not the rendered page.
- Check extractable structure. Clear headings, a direct answer near the top, and schema markup make your content easy to lift into an answer.
- Check off-site corroboration. Assistants decide who belongs in a category partly from who the wider web names alongside whom. If third-party sources never mention you next to your peers, you stay outside the set.
The tooling spans monitoring, digital PR, and integrated systems. Below is how the categories differ.
| Category | Examples | What it does | Best for |
|---|---|---|---|
| AI-mention monitoring | Otterly.ai, Profound | Tracks whether AI answers name your brand | Tracking whether AI answers name your brand |
| Digital PR / feature earning | BrandFeatured | Works to earn brand mentions and features in AI answers | Earning brand mentions and features in AI answers |
| SEO/content suite | Semrush, Conductor | Traditional search and content workflow | Traditional search and content workflow |
| Presence engine | CLEO | Measures AI visibility and improves readability, content, and corroboration | Measuring AI visibility and fixing readability, content, and corroboration in one place |
How to build the prompt set that makes the gap measurable
Step one above says "run your category's real prompts". That instruction fails in practice because most teams write prompts that flatter them. A prompt set that includes your brand name tells you nothing - assistants will discuss a brand you name. The set has to be built the way a buyer who has never heard of you would ask.
Build 15 prompts across five intents, none of which contain your brand name:
| Intent | Prompt shape | What its answer tells you |
|---|---|---|
| Category discovery | "What tools do teams use for [job to be done]?" | Whether you are in the category set at all |
| Ranked recommendation | "What is the best [category] for [segment]?" | Whether you are in the shortlist, and in what order |
| Constrained choice | "Best [category] for a [team size] with a [budget] budget" | Whether your positioning is understood well enough to match a constraint |
| Comparison | "[Named competitor] vs alternatives" | Whether you surface as an alternative to rivals |
| Problem-first | "How do I fix [the problem you solve]?" | Whether you are attached to the problem or only to the category label |
Run each prompt three times in a fresh session per engine. Three runs matter: assistant answers vary between sessions, and a brand that appears once in three is a different situation from one that appears three times in three. Record for every run: which brands were named, in what order, and whether a URL was attached. That last column is the one teams skip and the one that matters most - being mentioned and being cited with a link are different outcomes with different fixes.
The output is a baseline table: your appearance rate out of 45 runs (15 prompts x 3), and the same figure for the three or four brands that keep appearing instead. If your rate is zero, the problem is entity-level - the model does not associate you with the category, and no amount of on-page editing fixes that. If your rate is non-zero but low, the problem is competitive weight, and the on-page work in steps two and three is worth doing.
Which of the four checks is actually your blocker?
The four steps are ordered by cost, not by likelihood. Work them in order and stop at the first one that fails, because fixing a later step while an earlier one is broken produces no movement.
| Check | How to test it in under an hour | Fails if | Who fixes it |
|---|---|---|---|
| 1. Gap confirmed | Run the 45-run baseline above | You cannot state your appearance rate as a number | Marketing |
| 2. Machine readability | curl your key pages and read the raw response, or View Source and search for a sentence you can see on screen | The sentence is absent from the served HTML | Engineering |
| 3. Extractable structure | Read only the first sentence under each heading | Those sentences do not answer the heading | Editorial |
| 4. Off-site corroboration | Search your brand name excluding your own domain; count independent pages that name you alongside a category peer | Fewer than a handful, or none that mention a peer | PR / community |
Step two is the one worth doing first despite being second in the list, because it is the cheapest to test and the most absolute in effect. AI crawlers read the HTML your server sends and do not execute JavaScript. If your content is client-rendered, your page is effectively blank to them no matter how good it is - and every hour spent on steps three and four is wasted until that is fixed. One curl settles it.
Step four is the one most teams underestimate. AirOps and Kevin Indig found that 85% of brand mentions in AI answers originate from third-party pages rather than brand-owned content. A brand with a flawless, perfectly readable site and no independent mentions will still lose to a weaker site that the wider web discusses. For a fuller treatment of that failure mode, see how to increase brand citations in AI assistants.
Which sources do AI engines cite for this question?
In CLEO's own GEO tracking (August 2026), the query "my brand never appears when people ask ChatGPT for recommendations in our category and I don't know where to start fixing it" was run against ChatGPT Search, Claude, Google AI Overviews, and Perplexity. Unlike tool-selection queries, this pool is not made of listicles - it is made of explainers.
| Cited source | Type | Typical position |
|---|---|---|
| searchengineland.com | Trade-press definitional guide | 1 |
| entrepreneur.com | Mainstream business press | 1 |
| wealwayselevate.com | Agency how-to | 1 |
| linkedin.com (practitioner post) | Social post | 1 |
| arxiv.org | Academic preprint | 2 |
| nav43.com, insightland.org, memetik.ai | Agency and vendor explainers | 2 |
| llmvlab.com, tripledart.com, newlightdigital.com, seoforgpt.io | Vendor and agency guides | 3 |
Two features of this pool are worth acting on. First, a LinkedIn post ranks alongside Search Engine Land and Entrepreneur. For "where do I start" questions, engines weight practitioner accounts of doing the thing as heavily as institutional explainers - which means the off-site work in step four is not only about press coverage, and a substantive post on a social surface is a legitimate route into the pool. Second, an arXiv preprint sits at position two. Questions phrased as problems attract sources with methodology, not marketing. A page that states a testable procedure and reports what it produced competes here; a page that asserts benefits does not.
Citation pools shift as engines re-crawl. This is an August 2026 snapshot, published so it can be checked.
How does CLEO help a brand that is invisible in ChatGPT?
CLEO measures AI visibility across ChatGPT Search, Google AI Overviews, Perplexity, and Claude through the GEO Score (8 metrics, reported with 95% Wilson-confidence intervals). It diagnoses the readability layer with the AI Readability Score across six signals, including crawler access and JavaScript rendering. Quill produces AI-legible content, and the Social layer plus Citation Loop build the off-site corroboration that moves a brand into the category set. When CLEO ran its own site through this loop, AI Readability rose 35 to 96 and GEO 14 to 50 in 30 days; for DisburseCloud, AI citation share rose from 17% to 67% in 90 days at 95% Wilson confidence.
Frequently asked questions
How do I confirm my brand is absent?
Run 10-15 buyer questions through ChatGPT, Perplexity, and Google AI Overviews and log which brands appear. CLEO's GEO Score automates this with confidence intervals so you can tell a real change from noise.
What's the most common reason?
Machine readability. AI crawlers read server HTML and skip JavaScript, so a client-rendered site can be invisible. Check the raw server response first - it is the cheapest, highest-leverage fix.
How does CLEO help?
It measures AI visibility across ChatGPT, Google AI Overviews, Perplexity, and Claude, diagnoses readability, produces AI-legible content with Quill, and builds off-site corroboration - measuring and closing the gap, not just reporting it.
Check your score. Enter a domain at regencleo.ai/scan - the engine reads what ChatGPT, Google AI Overviews, and Perplexity say in response to your category's most-asked questions, reads where you sit in classical search, and reads what the social surfaces carry. It returns a single page, no login required. Paid GEO plans extend coverage to Claude.
