Does ChatGPT use my chat history when deciding which brands to mention?

It can when you test from a personal account. Saved memories, custom instructions and earlier conversations about your brand can all shape what a session returns, which skews a measurement in either direction. For baseline runs, use a dedicated testing account with nothing saved, or a temporary chat with memory and instructions switched off. Note in your log which setup each run used, so later comparisons stay like for like.

How many prompts should I test before trusting a zero-mention result?

Use a working threshold rather than a single pass: a fixed set of a couple of dozen prompts covering category, comparison and alternatives intent, run in full on several separate days. Call the brand absent only if it appears in none of those runs. The threshold is practical, not a statistical guarantee. If your category is broad, add prompts per sub-segment instead of rerunning the same few prompts more often.

What is the difference between a mention and a citation in ChatGPT?

A mention is your brand named in the answer text. A citation is a linked source the answer relies on, shown as a link or a source card. You can have one without the other: a model may name you from training data without linking you, or cite your page for a fact without naming your brand as a recommendation. Measure both separately and report them as distinct rates.

Do ads in ChatGPT change which brands it mentions in answers?

Treat paid placement and organic mentions as separate measurements. We could not verify OpenAI's advertising documentation when updating this article, so we make no claim about how ads interact with answers. In practice, exclude any sponsored or labelled unit from your baseline, since it is not part of the organic answer you track. Ask any vendor promising guaranteed mentions which documented OpenAI mechanism delivers them.

Do backlinks and domain authority still matter for getting mentioned in ChatGPT?

They still matter, but treat them as one input among several. An Ahrefs study of Google AI Overviews found that branded web mentions correlated more strongly with brand visibility than backlinks did. That is a correlation on a different AI surface, not proof of cause in ChatGPT. Keep earning links for the pages search systems retrieve, and pursue unlinked mentions in reviews, directories and editorial coverage as a separate effort.

Why Your Brand Gets Zero Mentions in ChatGPT

A zero-mention result in ChatGPT is either a measurement artefact (run-to-run variance, too few prompts, memory or custom instructions) or one of four real causes: blocked retrieval, a thin entity footprint, prompts that don't match how ChatGPT names brands, or no third-party corroboration. Rule out the artefacts first, then fix the causes in order.

What a zero-mention result actually measures

Start with a distinction worth drawing precisely. A mention is your brand named in the answer text. A citation is a linked source attached to the answer. The two signals move independently, so a zero can apply to one and not the other: a brand can be named without being linked, or linked as a source without being named. For a longer treatment, see how a mention differs from a citation in AI answers.

A mention rate only exists relative to a defined prompt set. "Zero mentions" means zero across the prompts you chose, on the dates you ran them, and nothing more. Track the same set over time, by hand in a spreadsheet or with a tool, so each result can be compared with the last.

ChatGPT answers through two paths. It can rely on what the model learned in training, or it can run a live web search and build the answer from retrieved pages. A zero can come from either path: the model may not know you, or search may not retrieve you.

Zeros are also common at the category level. BrightEdge reported on 14 August 2025, in its AI Catalyst research on ChatGPT, that "44% of prompts = zero brand mentions". A working rule follows. If competitors are also absent from the same answers and demand for the category is low, the zero is expected and says little about you. If competitors are named in the answers that omit you, the zero is a finding.

A real zero also costs more than it used to. Profound reported on 19 May 2026 that, after ChatGPT began turning brand mentions into links, "Homepages went from a rounding error in OpenAI referral mix to roughly 1 in 4 clicks in a week." A mention is now a referral path as well as a visibility signal, so a missing mention also means missing traffic.

Rule out a measurement artefact first

Only diagnose a zero that survives clean, repeated runs. The protocol below takes a few hours and prevents weeks spent fixing a problem that does not exist.

Run-to-run variance and prompt-set size

Manual prompt testing is the baseline. Mix category prompts ("best accounting software for freelancers"), comparison prompts ("Brand A vs Brand B for small agencies") and alternatives prompts ("alternatives to Brand A"). Use the wording your buyers would use, not your own positioning lines.

The same prompt returns different brands from one run to the next, so a single pass proves nothing. Before calling a brand absent, apply a working threshold rather than a statistical guarantee: a fixed set of a couple of dozen prompts, several complete runs on separate days, and zero mentions across all of them. A brand that appears in some runs but not others is present but unstable, which is a different problem from absence.

Memory, custom instructions and clean-state testing

ChatGPT can personalise answers through saved memories and custom instructions. An account that has discussed your brand, or that describes your company in its instructions, is a contaminated test environment. Test in a clean state: a temporary chat, memory off, no custom instructions. Keep only clean-state results in the baseline, and discard any run where you are unsure of the settings. These features change, so check OpenAI's Help Center for their current behaviour before each testing round.

Account tier, model and web search: record them with every run

For every run, log the account tier, the model selected and whether a web search fired, which usually shows as linked sources in the answer. This is a testing method, not a claim that tiers behave differently: the log lets you tell a brand problem apart from a configuration difference. If zeros cluster on runs without web search, the training path is the likelier gap. If they persist when search fires, look at retrieval and corroboration.

The four real causes, in diagnosis order

Work through the causes from cheapest to verify to slowest to repair. Stop at the first one that confirms, fix it, then re-measure before moving on. If competitors keep appearing where you don't, read why competitors get cited by ChatGPT when you aren't for the comparison side of the diagnosis.

Cause	How to confirm it	Fix
ChatGPT cannot retrieve your site	robots.txt blocks OAI-SearchBot or ChatGPT-User, or logs show no hits from them	Allow the search and user agents, and check CDN or firewall rules
No entity footprint to reuse	ChatGPT describes you vaguely or wrongly, or confuses you with a similar name	Clear entity pages, consistent profile links, current facts
Prompts don't match how ChatGPT names brands	Everyone is absent on informational prompts, and brands appear on buying prompts	Rebalance the set toward transactional intent and publish pages for those questions
No third-party corroboration	Competitors appear on review sites, directories and lists where you are absent	Earn reviews, listings and editorial mentions

First cause: ChatGPT cannot retrieve your site

OpenAI's crawler overview, verified on 19 October 2026, lists separate user agents: GPTBot for model training, OAI-SearchBot for ChatGPT search, and ChatGPT-User for visits triggered by a user's request. The documentation states: "Each setting is independent of the others – for example, a webmaster can allow OAI-SearchBot in order to appear in search results while disallowing GPTBot". Blocking OAI-SearchBot can keep you out of search-based answers even when GPTBot is allowed.

A robots.txt that stays visible in ChatGPT search while opting out of training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: GPTBot
Disallow: /

Adjust the GPTBot group to your own training policy. robots.txt states intent; server logs show behaviour. Search your access logs for all three agents, for example with grep -E "GPTBot|OAI-SearchBot|ChatGPT-User" access.log, and read the status codes. Forbidden or unavailable responses usually point to a CDN, firewall or bot-management rule that blocks the agents whatever robots.txt says. No OAI-SearchBot hits at all over several weeks suggests ChatGPT search is not reaching you.

Second cause: no entity footprint to reuse

If retrieval works, check what ChatGPT can say about you. Structure entity pages, such as home, about, product and pricing, so a model can extract a clear answer: what the company is, who it serves, where it operates and what it costs. Put those facts in plain sentences near the top of the page, and keep brand information current and accurate everywhere it appears.

Then run a disambiguation check. In a clean state, ask ChatGPT what your brand is and who runs it. If it confuses you with a similarly named company, align the signals it uses to tell you apart: the same name, the same one-line description and sameAs-style links between your site, your LinkedIn page and your directory profiles. Organization schema with sameAs properties expresses the same links in structured form.

Freshness belongs to this cause too. Ahrefs reported on 28 July 2025 that "The average age of URLs cited by AI assistants is 1064 days, compared to 1432 days for URLs in organic SERPs—25.7% 'fresher'." Pages with stale facts give a retrieval-based answer less reason to choose them. Update core pages when prices, features or locations change, and show the update date.

Third cause: your prompts don't match how ChatGPT names brands

Some zeros come from the prompt set, not the brand. BrightEdge reported on 14 August 2025 that "Transactional queries outpace informational by 8–10x" for brand mentions in ChatGPT. A set built mostly from informational questions, such as "how does invoicing work", under-counts mentions for every brand, including yours. Split results by intent to confirm. If you are absent on buying prompts where competitors appear, the cause lies elsewhere. If every brand is absent on informational prompts, rebalance the set toward comparison, alternatives and "best for" prompts, and publish pages that answer those buying questions directly.

Fourth cause: no third-party corroboration

The slowest cause to repair is a lack of independent sources naming you. Treat third-party mentions as a lever distinct from backlinks: an unlinked mention in a review, a comparison list or a trade article still shows that others recognise the brand. Reviews and review platforms count. Outside software, the same role falls to local surfaces: Google Business Profile reviews and industry directory listings for a local business.

On links versus mentions, Ahrefs reported on 26 May 2025: "Web mentions (0.664) correlate much more strongly than backlinks (0.218)." Read that with its scope. The study measured brand visibility in Google's AI Overviews, not in ChatGPT, and it measured correlation, not causation. Backlinks and domain authority plausibly still matter for the pages search systems retrieve, but in this data, branded mentions across the web track AI visibility more closely.

Ads in ChatGPT and paid placement in answers

Advertising is a separate question from organic mentions. When we prepared this update on 19 October 2026, we could not load OpenAI's advertising documentation to verify how ads are shown, labelled or kept apart from answers, so this section states no policy on OpenAI's behalf. Read OpenAI's current advertising pages directly before relying on any claim about them, including claims made by tool vendors.

Whatever the policy says, keep paid placements out of your data. A labelled ad is not an organic mention: if a test answer includes a sponsored or labelled unit, exclude it from the baseline, because the placement is not part of the organic answer you are measuring. When a vendor promises guaranteed mentions in ChatGPT answers, ask which documented mechanism delivers them. If no public documentation describes that mechanism, treat the promise as unverified.

Where to start this week

Freeze a prompt set: a couple of dozen category, comparison and alternatives prompts in your buyers' wording.
Run it in a clean state on several separate days, and record account tier, model and web-search status with each run.
Check retrieval: read robots.txt for GPTBot, OAI-SearchBot and ChatGPT-User, then search your server logs for all three and their response codes.
Check the entity: ask ChatGPT what your brand is, fix any confusion with a similarly named company, and refresh stale facts on core pages.
Check the prompts: split results by intent and add transactional prompts if the set leans informational.
Check corroboration: list the review sites, directories and comparison pages that name your competitors, and start where you are missing.
Pick a re-measure date a few weeks out and re-run the same fixed prompt set after each fix, so every change stays attributable. To repeat runs on a schedule, see how to automate AI visibility checks.

Run a free Namedrop scan to see which of your prompts already mention your brand and where your zeros come from.