Model: ChatGPT · Custom Instructions
On 15 July 2026, OpenAI raised ChatGPT’s custom instructions limit to 5,000 characters — more than triple the previous 1,500 — giving Plus, Pro, Enterprise, Business and Education users far more room to shape how it answers. Here’s what a bigger personalisation window changes for whether AI names your brand.
Source: ChatGPT release notes — Increased custom instructions limit (15 July 2026)
- On 15 July 2026, ChatGPT increased its custom instructions limit to 5,000 characters, up from 1,500.
- It applies to Plus, Pro, Enterprise, Business and Education users.
- Custom instructions are standing personalisation: a persistent brief that shapes every answer’s style and behaviour.
- More personalisation means AI recommendations vary more from person to person — the same question can surface different brands.
- Personalised answers are harder to track, so AI visibility has to be measured across contexts, not from a single logged-out prompt.
What OpenAI actually changed
On 15 July 2026, OpenAI updated its release notes to confirm it is “increasing the character limit for custom instructions in ChatGPT.” Plus, Pro, Enterprise, Business and Education users can now save up to 5,000 characters, up from 1,500 — more than triple the old ceiling — giving them far more room to customise ChatGPT’s response style and behaviour.
Custom instructions are the standing brief a user gives ChatGPT once and reuses on every chat: who they are, what they care about, and how they want answers framed. Unlike a one-off prompt, this context persists in the background of each conversation. Raising the limit does not add a new model or a new answer surface — it deepens how much a person can bend the model toward their own preferences, every single time they ask.
Why personalisation shapes AI answers
Every answer ChatGPT gives is a blend of the live web, the model’s general knowledge, and whatever it knows about the person asking. Custom instructions feed straight into that last part. When a user writes “I run a small B2B software firm and prefer tools with a free tier,” ChatGPT weighs its shortlist against that brief — and the brands it names shift accordingly. The ranking logic behind that pick is the same one we set out in how LLMs decide which brands to cite.
This is the same lever we examined with ChatGPT memory sources: the more a model knows about someone, the more its recommendations bend to fit them. Custom instructions are the explicit, user-authored version of that, and a 5,000-character brief carries a lot more signal than a 1,500-character one. For brands, it means an AI recommendation is increasingly a conversation between your public evidence and a reader’s stated preferences.
What a 5,000-character window unlocks
At 1,500 characters, most people could fit a role and a couple of preferences. At 5,000, a user can describe their industry, their constraints, their budget, the tools they already use, the outcomes they want, and the tone they expect — a full working profile. That richer brief gives ChatGPT far more to match against when it decides which products, services or sources to put forward, echoing the sharper recommendation behaviour we covered when GPT-5.5 Instant levelled up recommendations.
The upshot is more specific, more opinionated answers. A generic “best CRM” question filtered through a detailed profile becomes “best CRM for a five-person agency on a tight budget that already uses Google Workspace” — and the shortlist narrows fast. Fitting that tighter brief is where visibility is won or lost.
What the bigger custom instructions limit changes
ChatGPT custom instructions expanded to 5,000 characters, live from 15 July 2026.
Does more personalisation help or hide you?
It cuts both ways. If your brand is a genuinely strong fit for a user’s stated needs, a detailed brief helps — it hands ChatGPT the exact criteria your product satisfies, and you become the obvious pick. If you are a weak or generic fit, the same detail filters you out faster than a broad question ever would. Personalisation rewards specificity of fit, which raises the bar set out in is your brand ready for AI search?
There is also a limit to how far custom instructions reach. When ChatGPT prioritises a quick reply, it can lean less on stored context and more on the live web and its base knowledge — the behaviour we flagged when it began bypassing memory for fast answers. So a strong public footprint still does the heavy lifting for the many queries where a personal brief is thin or set aside. The practical read: personalisation amplifies fit, it does not replace evidence.
Personalised answers are harder to track
The measurement problem is the real story here. If two users ask the same question and get different shortlists because their custom instructions differ, there is no single “correct” answer to check. A logged-out prompt tells you one version of reality; a customised account tells you another. That variability is the same challenge a model-picker change introduces, and it grows as personalisation deepens.
It also compounds with every new place ChatGPT answers. The redesigned ChatGPT desktop app carries a signed-in user’s custom instructions into work and chat modes alike, so the personalised shortlist follows them across surfaces. Tracking your presence from one anonymous query no longer captures how most people actually experience the answer.
What this means for your content
The response is not to chase individual profiles — you cannot see them — but to make your fit unmistakable in public. Spell out who your product is for, the constraints it suits, the use cases it wins, and the outcomes it delivers, in clear content a model can read and reuse. That is the groundwork we keep returning to in building a GEO strategy: the more precisely your evidence maps to real buying criteria, the more custom-instruction briefs it will match.
This mirrors what platforms reward elsewhere. Google’s push on preferred sources for AI Overviews favours clear, trustworthy pages, and the same discipline serves you inside ChatGPT. If your category can be reached commercially, our overview of ChatGPT advertising covers where paid placement is heading as these answers get more personal.
How to check your ChatGPT visibility
Because personalisation now carries more weight, one prompt is no longer a fair test. Here is a short, practical way to see how ChatGPT names your brand across contexts:
- Baseline logged out. Run your category’s buying questions with no account and record the shortlist — today is 20 July 2026.
- Add a realistic profile. In a signed-in account, write custom instructions that match a typical customer, then ask the same questions again.
- Vary the brief. Change the budget, industry or constraints and watch how the named brands move.
- Test a strong-fit case. Write instructions that describe your ideal buyer and check whether you appear when you clearly should.
- Test a weak-fit case. Describe a poor-fit buyer and confirm the answer is filtering sensibly, not against you unfairly.
- Repeat over time. Personalisation and models both change, so one run is a snapshot; the value is in the trend.
How reconnAI tracks personalised answers
reconnAI monitors how the leading AI models — ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — answer questions across regions and contexts, and re-baselines that tracking whenever a platform changes how its answers are shaped. A bigger custom instructions window is exactly that kind of change: it widens the gap between a generic answer and a personalised one, and that gap is where brands quietly gain or lose ground.
It joins a busy run of shifts we have tracked lately — from ChatGPT Work and ChatGPT Sites to ChatGPT on WhatsApp, Perplexity Brain and Claude Sonnet 5. If you want to see how ChatGPT represents your brand as personalisation deepens, get in touch with our team or explore how AI visibility tracking works.
ChatGPT custom instructions: your questions answered
What changed with ChatGPT custom instructions? On 15 July 2026 OpenAI raised the character limit from 1,500 to 5,000 for Plus, Pro, Enterprise, Business and Education users, giving them more room to customise ChatGPT’s response style and behaviour.
Are custom instructions the same as memory? Not quite. Custom instructions are a standing brief you write yourself; memory is context ChatGPT gathers over time. Both personalise answers, and both can change which brands a reply names.
Does this change which brands ChatGPT recommends? It can. A richer profile lets ChatGPT match its shortlist more tightly to a user’s stated needs, so strong-fit brands surface more and generic ones fade. Public evidence of fit is what wins that match.
How do I track my brand across personalised answers? Compare logged-out and signed-in results, vary the custom instructions to reflect different customers, and watch the trend over time rather than relying on a single prompt.
About reconnAI
reconnAI tracks how the major AI models represent topics and sources across ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — across multiple regions. We monitor how those models answer and how they change over time, so you can stay ahead of shifts in the AI landscape.
Last updated: 20 July 2026.