How to get ChatGPT to recommend your brand: what actually works

You cannot pay ChatGPT to recommend your business, and there is no ranking ladder to climb, but you can earn a place in its answers. ChatGPT's answers are probabilistic rather than ranked: SparkToro research cited by Ahrefs found less than a 1-in-100 chance that ChatGPT gives the same list of brands twice for the same query. The commercial stakes are real all the same: G2's 2025 Buyer Behavior Report, reported by HubSpot, ranks "generative AI chatbots" as the number-one influence on B2B supplier shortlists, ahead of review sites, supplier websites and salespeople. So the realistic goal is not to rank number one on ChatGPT. It is to increase the share of buyer questions in your category where ChatGPT names your brand at all. This guide covers seven tactics with published evidence behind them, and the myths that waste budget.

At a glance: the seven tactics

How does ChatGPT decide which brands to recommend?

ChatGPT recommends brands from two inputs: what its training data contains about your category, and, for buying questions, what it finds when it searches the live web while composing the answer. According to a Nectiv analysis from October 2025 cited by HubSpot, commercial-intent prompts trigger web searches in ChatGPT 53.5% of the time, versus 18.7% for informational queries. So the answer to "best project management tool for a 10-person agency that bills hourly" is assembled largely from live sources (review sites, comparison articles, forums) retrieved at the moment of asking.

The full mechanics deserve their own piece: see our explainer on how AI assistants decide which brands to recommend. For this guide, the practical takeaway is that you influence ChatGPT's recommendations by influencing the sources it retrieves, not by optimising a hidden ranking factor.

1. Measure where you're absent before you spend anything

Start by measuring which buyer questions in your category already name your brand and which do not. Effort spent where you are already visible is wasted, and effort spent blind is guesswork. This is what AI visibility is: the share of realistic buyer questions where an AI assistant names your brand, measured against who it names instead.

The questions that matter most are the ones that never mention you: non-branded questions, in the vocabulary we use throughout this guide. If a buyer asks "is Thameside Projects any good?", they already know the brand exists. If they ask "best project management tool for a 10-person agency that bills hourly", the winner is whoever the assistant names unprompted. Appearing only when a buyer already types your name is not visibility.

You can measure this by hand: write 20–25 realistic buyer questions for your category, run them through ChatGPT in fresh sessions, and log every brand and source named. Our guide on how to measure your AI visibility step by step sets out that protocol. Or run a free AI visibility check with Skavora: it researches your brand live, generates 25 conversational buyer questions (around 80% of them non-branded), runs them through Gemini with live web grounding (the engine searching the web as it answers, rather than relying on training data alone) and returns a scored report with the competitor leaderboard and the cited-source breakdown that turns the rest of this guide into a to-do list.

Either way, by the end of this step you should know:

• Which of your category's buyer questions name you, and which do not
• Who gets named in the gaps
• Which sources those answers cite

A caveat for any method, ours included: live answers vary between runs, so treat one measurement as a starting position to re-check, not a fixed score.

2. Be present on the sources ChatGPT actually cites

The pages ChatGPT cites are largely not the pages that rank in Google, so a strong Google presence does not carry over automatically. An Ahrefs study of 15,000 prompts found only 12% of URLs cited by ChatGPT, Gemini and Copilot appear in Google's top 10 for the same query. Separately, Ahrefs found that roughly 80% of citations from large language models (LLMs, the models behind assistants like ChatGPT, Gemini and Claude) come from pages that do not rank in Google's top 100 at all (both via ZipTie). Seer Interactive found 28% of ChatGPT's most-cited pages have zero Google organic visibility (via Onely).

The implication is to identify the specific sites cited for your category's questions and earn presence there. The cited-source mix varies sharply by industry, so a scan of your own category tells you which sources AI engines cite in your category rather than someone else's. The recurring types:

• Review and comparison sites (G2, Trustpilot, Capterra, category-specific equivalents)
• Reddit threads and specialist forums
• YouTube reviews and walkthroughs
• Editorial "best of" round-ups in trade press
• Reference sites and industry directories

Treat that list as an outreach plan, not a reading list: an earned mention on a source ChatGPT already cites has a route into answers that a page on a site it never retrieves does not.

3. Earn reviews and get into comparison content

"Best of" list posts are the single largest source of ChatGPT's citations on commercial queries. Research by Glen Allsopp, cited by Ahrefs, found they account for roughly 43.83% of ChatGPT citations for commercial questions. When ChatGPT answers "best project management tool for a 10-person agency that bills hourly", it is largely synthesising existing round-ups and review profiles. If you are absent from those, you are absent from the answer.

Practical moves, in order of effort:

• Claim and complete your profiles on the review platforms your category uses. An empty G2 or Trustpilot page is a missed citation.
• Ask real customers for reviews, steadily rather than in bursts.
• Pitch the round-ups that already exist. Find your category's "best X for Y" articles and email the authors a genuine case for inclusion; those pages are updated regularly and their authors need credible candidates.
• Publish honest comparison pages yourself, your product versus named alternatives, with a real feature table. A fair, structured comparison is the shape engines lift.

4. Build a Reddit and community presence, honestly

Presence across multiple independent communities is one of the strongest measured correlates of AI visibility. An Evertune study of 75,000 brands found brands present on four or more non-affiliated forums are 2.8x more likely to appear in ChatGPT responses, and a Search Engine Land study of 800+ websites across 11 industries found multi-platform brand mentions correlate with AI visibility at r=0.87, a strong association on a scale where 1.0 would be a perfect match (both via ZipTie). Both findings are correlations: they show that visible brands tend to be present in many places, not that posting in forums causes citations.

"Honestly" is doing real work in that heading. Astroturfing (fake accounts praising your own product) breaks every major platform's rules, is visible to moderators, and is increasingly detectable in aggregate. The durable version:

• Answer questions in your category's subreddits and forums as yourself, disclosed, mentioning your product only when it genuinely answers the question
• Encourage satisfied customers to share their experience where they already post
• Correct factual errors about your product politely, where your category is discussed
• Accept that many contributions will not mention you at all: presence and credibility compound

5. Structure your content so an AI can lift it

Text that works as a complete answer on its own is what AI engines quote, and the citation evidence on this is unusually specific. A Kevin Indig analysis of 1.2 million ChatGPT citations, summarised by Ahrefs, found cited text is nearly twice as likely to contain definitive phrases such as "is defined as" (36.2% vs 20.2%), that question marks in headings roughly double citation rate (18% vs 8.9%), and that 44.2% of citations come from the first 30% of a page. Ekamoira found self-contained 50–150-word chunks earn 2.3x more citations, and the Princeton/Georgia Tech GEO paper (ACM KDD 2024) found that adding statistics lifted AI-answer visibility 15–40% (both via ZipTie).

Read those carefully: the Indig figures describe what cited text looks like, which is a pattern rather than proof that reformatting a page causes citations. The KDD paper is the exception: it tested changes directly and measured the lift.

In practice:

• Front-load the answer: the conclusion goes in the first sentence of the page and of each section
• Phrase headings as the questions buyers actually type
• Keep paragraphs self-contained, no "as mentioned above", no pronouns pointing at the previous section
• Name entities explicitly: products, platforms, prices, places

A contrast pair shows the difference. Both passages below are invented examples of writing style: "Thameside Projects" is not a real product and the figures in it are not findings.

Hard to extract: "As mentioned above, there are several factors to consider here, and our approach to this has evolved considerably over the years. We believe this makes a real difference for teams like the ones we discussed."

Easy to extract: "Thameside Projects is a project management tool for agencies of 5–20 people that bill by the hour. It includes built-in time tracking, per-client budgets and invoice export, and pricing starts at £8 per user per month."

The second version names the entity, states the facts, and depends on nothing around it. That is the sentence shape AI answers are built from.

6. Keep your entity facts consistent everywhere

Give every source that describes your brand the same facts, so an engine assembling them gets one clear picture rather than three conflicting ones. If your website says one thing about what you do, your LinkedIn another and your directory listings a third, an assistant synthesising them has less to work with than it would for a brand described identically everywhere. Be clear on the status of this tactic: none of the studies cited in this guide measures naming consistency directly, so treat it as cheap housekeeping rather than a proven lever.

The fix is unglamorous consistency:

• One canonical description of what you do, who it is for and what it costs, reused verbatim across your site, social profiles, directories and review platforms
• Consistent naming: one brand name, spelled one way, everywhere
• An about page stating plainly what the company is, where it operates and since when
• Structured data (Organization schema) on your site confirming the same facts

It is a short job, and the same entity picture feeds Gemini, ChatGPT, Claude and Google AI Overviews alike.

7. Keep your content genuinely fresh

Recency shows up strongly in the citation data. An AirOps study cited by Semrush found 95% of ChatGPT citations come from content published or updated within the last 10 months, and that pages with a visible "last updated" timestamp receive 1.8x more citations. Ahrefs' 17-million-citation study across seven AI platforms found AI assistants prefer content that is 25.7% fresher, with ChatGPT citing URLs 393 days newer than Google's organic results in text and 458 days newer in its citations (Ahrefs).

Genuine freshness means:

• Updating the substance (numbers, examples, findings) on a real cadence
• Showing the date on the page as well as in the markup
• Refreshing the same URL rather than republishing at a new one

Bumping a timestamp without changing the page is theatre, and leaves the content stale for the next retrieval.

What doesn't work: three myths about ranking on ChatGPT

Can you pay ChatGPT to recommend your business?

No. OpenAI's own documentation states that ChatGPT product results are "organic and unsponsored, ranked purely on relevance" (via HubSpot): there is no placement to buy in those results, and no agency or partner can insert your brand into them. Anyone selling guaranteed placement in ChatGPT's recommendations is selling an outcome they do not control. This reflects OpenAI's published position as of July 2026; platform commercial models change, so check the primary source before relying on it.

Can an agency guarantee ChatGPT will recommend your brand?

No, and the SparkToro finding explains why: with less than a 1-in-100 chance of the same brand list appearing twice (via Ahrefs), nobody can honestly promise a specific answer to a specific prompt. What can be improved, and measured, is the share of your category's buyer questions where you appear. Judge any effort (ours included) on that number, re-measured over time, rather than on a screenshot of one lucky answer.

Is it enough to win on one source, like Reddit?

No, because citation patterns move at the source level. A Semrush study tracking 230,000 prompts across 13 weekly snapshots found ChatGPT cited Reddit in roughly 60% of responses in early August 2025, falling to around 10% by mid-September, while Wikipedia dropped from about 55% to under 20% (Semrush). A strategy built entirely on one platform inherits that platform's volatility. The durable position is the boring one: genuine presence across reviews, communities, comparisons and your own well-structured site, so no single source's collapse takes you with it.

The short version: earn presence where ChatGPT already looks

• Measure first: find the non-branded buyer questions where you are absent, and who fills the gap
• Earn presence on the sources ChatGPT cites in your category, mostly not your Google competitors
• Get into reviews and "best of" comparisons; they account for the largest share of commercial citations
• Take part in Reddit and forums as yourself, disclosed, usefully
• Write extractable pages: question headings, front-loaded answers, self-contained paragraphs
• Keep entity facts consistent and content genuinely fresh, then re-measure

None of this is a trick, and none of it can be faked for long. None of it guarantees a result either: no one controls what an AI engine says, so the honest target is a measured improvement in how often you are named, not a promised placement. For the starting measurement, run a free Skavora scan and get the scored report, competitor leaderboard and cited-source outreach list in one pass.