When the chatbot keeps naming someone else
You ask ChatGPT for the best option in your category and it recommends a competitor. You try Perplexity and get the same name. This is not a ranking you can chase with a title tag, because there is no page one to climb. The model is assembling an answer from sources it trusts, and right now those sources point at your rival instead of you.
This guide is a method, not a tool tour. It covers how to diagnose why a specific competitor wins the AI answer, how to find the exact citation gaps between you and them, and the plays that actually move the answer in your favour. For the general monitoring process, see how to monitor AI brand mentions. For the share-of-voice maths, see how to measure AI share of voice. Here we stay focused on one thing: overtaking a named competitor.
The numbers behind who the chatbot picks
| What the data shows | Figure | Source | Why it matters when you are chasing a rival |
|---|---|---|---|
| The same query changes its answer | roughly 70% of the time | SparkToro | One screenshot of a rival winning proves nothing, measure a distribution |
| Two identical queries return the same brands | under a 1-in-100 chance | SparkToro | Judge who wins across repeated runs, not a single prompt |
| Reddit's share of all AI citations | 40.1% | Semrush | A community thread naming your rival can outweigh your own pages |
| Citations that never name the brand | 62% | Semrush | Rivals can feed the answer invisibly, so compete for unnamed sources too |
| What AI visibility tracks to most | third-party mentions and video | Ahrefs | The lever is off-site sources, not your homepage |
First, accept that the answer is a moving target
Before you diagnose anything, understand what you are up against. The same AI query changes its answer roughly 70% of the time, and the odds of two identical queries returning the same brand list are under 1 in 100 (SparkToro). So a single screenshot of the chatbot praising your competitor is not proof they have won. It is one roll of the dice.
This matters for how you diagnose. You cannot judge who wins from one prompt. You need the same question asked repeatedly, across models, over days, so you are measuring a distribution and not a fluke. A rival who appears in 8 of 10 runs is genuinely beating you. A rival who appears in 2 of 10 is beatable this quarter.
Step 1: Find out which sources cite your competitor
AI answers are built on citations. The fastest way to understand why a competitor wins is to read the sources the model leans on when it names them. This is easier on some engines than others: ChatGPT and Gemini often hide their citations, while Perplexity and Claude expose theirs consistently. Start your diagnosis on Perplexity and Claude, because they hand you the source list directly.
Run your core buying question on Perplexity and Claude, then open every cited link. You are looking for three things: which domains appear, whether your competitor is named on those pages, and whether you are absent from the same pages. Log the cited URL, the domain type (review site, listicle, forum thread, video, news), and whether each brand is present. That log is your battlefield map.
The sources that decide who gets named
Not all citations carry equal weight, and community content dominates. Reddit alone is 40.1% of all AI citations, the single most-cited source across engines (Semrush). That means a Reddit thread where your competitor is recommended can outweigh a dozen of your own landing pages.
Share of AI citations
Most-cited domains in AI answers
There is a catch that works in your favour. 62% of AI citations never name the brand behind them (Semrush). The model pulls a fact or a recommendation from a page without attributing it to any company. So a competitor can be feeding the answer through review sites and forum threads without their name ever appearing in the sources you can see. Do not assume the visible citations tell the whole story. If a source keeps appearing and no brand is named, that source is still shaping the answer, and it is a place you can compete.
Step 2: Map the citation gap
Now turn your source log into a gap list. For every cited source where your competitor appears and you do not, you have a specific, addressable gap. Rank them by how often each source shows up across your repeated runs. A review site cited in 7 of 10 answers is a priority. A stray blog cited once is not.
Want to see this in action?
See how every major AI model talks about your brand. Free to start.
Then split the gaps into the three things AI visibility actually correlates with. Ahrefs found that AI visibility correlates most strongly with third-party mentions and video, not on-page work. So sort your gaps into: third-party mentions (review sites, listicles, news), community (forums and Reddit threads), and video (YouTube and embedded clips). If most of your competitor's edge sits in third-party mentions and community, that tells you where to spend, and it tells you that rewriting your own homepage will not close the gap.
Why competitors win, and what to do about it
| Why the competitor wins the answer | What it looks like in your citation log | Your play to overtake |
|---|---|---|
| They are recommended in the top-cited review sites and listicles | Their name on the same third-party pages the model cites; you are absent | Earn a place on those exact pages: pitch inclusion, correct outdated entries, get customers to leave verified reviews |
| They own the community threads | Reddit and forum links in the citations name them, or answer the question they rank for | Show up authentically in the threads the model cites; answer the real question, do not spam a link |
| They are cited through video | YouTube or embedded clips appear in the sources | Publish or sponsor video that answers the buying question, since video correlates with AI visibility (Ahrefs) |
| They feed the answer without being named | A source recurs, recommends a product, but no brand is shown (62% of citations name no brand, per Semrush) | Compete for that unnamed source anyway: it is shaping the answer even when attribution is missing |
| They simply appear more often across runs | They show in 8 of 10 repeated queries, you show in 2 | Widen coverage across all the above so your presence compounds, then re-measure the distribution |
Step 3: Earn presence in the sources that already win
You have the ranked gap list. Start at the top. If a review site or listicle is cited in most of your competitor's answers, your job is to get onto that specific page, not to build a new one. That means pitching the editor for inclusion, correcting stale or wrong entries about your category, and driving verified customer reviews so the page has a reason to mention you. This is slow, unglamorous work, and it is the work that moves the answer, because the model already trusts those pages.
Do not waste the cycle rewriting your own site copy first. On-page work is the weakest lever here (Ahrefs). Your landing page matters for conversion once the chatbot sends someone, but it is rarely what tips the recommendation. Fix the sources the model cites before you touch your own meta descriptions.
Step 4: Third-party mentions and community
Community is where the largest share of citations lives, so it deserves a deliberate play rather than an afterthought. Reddit is 40.1% of all AI citations (Semrush), which makes the threads in your niche disproportionately powerful. Find the threads the model already cites, then show up as a real participant: answer the question genuinely, share specifics, and let your product come up because it fits, not because you dropped a link.
One warning about betting everything on a single source. Reddit's share of ChatGPT citations fell from around 60% to around 10% in a fortnight in late 2025 (Semrush). Engines re-weight their sources without notice. So spread your effort across community, review sites, and video rather than pouring it all into one platform. Diversity of mentions is what survives a re-weighting.
Step 5: Track the overtake, do not eyeball it
Because a single query is a coin toss (SparkToro), you cannot tell whether your work is landing by checking the chatbot once a week. You need the same buying questions run daily across models, with the competitor tracked beside you, so you can watch your share of the answer rise as theirs falls. That is the only honest read on whether you are overtaking them.
This is the category the measurement tools serve. To disclose our own: Honeyb (ours) checks all four major models daily and reports sentiment, citations, competitor tracking and recommendations, with a free check to start. Others solve pieces of it: Otterly AI starts at $29/mo with a free trial, Peec AI from around $89/mo with multilingual coverage, SE Ranking folds AI visibility into an SEO suite from around $55/mo, and Profound sits at the enterprise end from around $399/mo rising into the thousands with deep analytics and API access. For a fair side-by-side, see the best AI visibility tools roundup and the AI search competitive analysis tools listicle, or the head-to-head Honeyb vs Profound comparison.
Put the method on repeat
Overtaking a competitor in AI answers is a loop, not a launch. Read the sources that cite them, rank the gaps, earn presence in the winning sources, build community and video coverage, then re-measure the distribution and repeat. The rival who looks unbeatable in one screenshot is usually beatable across ten runs, once you compete on the sources the model actually reads. Agencies running this for several clients can systematise it via Honeyb for agencies, and teams tracking sentiment alongside share can layer in AI sentiment tracking.
Start by seeing where you stand against your rival today. Run a free AI visibility check to find out which sources are naming your competitor and where your gaps are.
