Updated June 2026. Originally published February 2023 — approximately ten weeks after ChatGPT's November 30, 2022 launch. Rebuilt as EPR's current reference on integrating large language models into communications strategy, with the substantial evolution the category has undergone across three and a half years.
Limitations to ChatGPT: What to Know Before Integrating AI Into Your Strategy
In February 2023, ChatGPT had been publicly available for approximately ten weeks. The product had reached 100 million users in the fastest consumer technology adoption cycle in recorded history. Communications professionals across PR, marketing, and corporate communications were rapidly experimenting with the product. The question framing the original piece — what should communications professionals understand about ChatGPT's limitations before integrating it into operational workflows — was urgent and substantially undeveloped at the time.
Three and a half years later, the AI category has evolved substantially. The original ChatGPT product (built on GPT-3.5) has been succeeded by GPT-4 (March 2023), GPT-4 Turbo (November 2023), GPT-4o (May 2024), GPT-5 (2025), and the broader contemporary model family. Multiple major competitors have emerged at sustained scale — Anthropic's Claude (the original 2023 release through Claude 3 in March 2024 through Claude Opus 4.x in 2026), Google's Gemini family, Meta's Llama family operating as the dominant open-weights infrastructure, xAI's Grok, and the broader Chinese AI model ecosystem (DeepSeek, Qwen, GLM, and others).
The communications integration of AI tools has matured substantially. Many of the limitations the original piece flagged still apply structurally. Others have been substantially addressed by subsequent model evolution. This page is EPR's current reference on what communications professionals should understand about AI tool integration in 2026.
What the Original 2023 Piece Flagged
The original February 2023 piece identified five limitations.
Context understanding. The piece noted that ChatGPT could generate authoritative-sounding content without genuinely understanding context, producing inaccuracies on complex or nuanced issues.
Data freshness. The piece flagged the September 2021 knowledge cutoff of the original ChatGPT and the implications for content covering current events.
Tone replication. The piece noted that ChatGPT could produce mechanical-feeling content that lacked the empathy and personal voice that real human communications produced.
Scalability. The piece flagged enterprise-scale deployment challenges.
Personal information handling. The piece noted the operational security concerns around inputting sensitive data into ChatGPT.
What Three and a Half Years Have Demonstrated
Three of the original five limitations remain structurally relevant. Two have been substantially addressed.
Context understanding remains a sustained challenge. Despite substantial model capability improvements, large language models continue to produce authoritative-sounding content with factual errors, missing context, and nuanced reasoning failures — particularly on topics requiring specialized expertise, recent events, or subjective judgment. The 2026 category continues to require human expert review for any content that will be published, sent to clients, or used in operational decision-making. The phenomenon of "AI hallucination" — confident generation of false information — remains substantially unsolved.
Data freshness has been substantially addressed. The 2023-era knowledge cutoff problem has been addressed through multiple mechanisms — retrieval-augmented generation (RAG), web search integration (ChatGPT, Claude, Perplexity, Gemini, and Grok all now operate with live web access), and the rolling updates to model training data. The original 2023 limitation has substantially evolved.
Tone replication has improved substantially but remains incomplete. Contemporary models produce substantially more natural-feeling content than 2023-era models. The discipline of voice and tone replication for specific brands and individuals has matured substantially. However, the structural limitation — that AI-generated content typically lacks the specific lived experience, judgment, and authentic voice of a human author — remains. Communications operations using AI for drafting consistently produce better outcomes when human writers edit and refine AI-generated drafts rather than publishing them directly.
Scalability has been substantially addressed through enterprise infrastructure. The 2023-era scalability concerns have been substantially addressed through enterprise API access at all major providers, dedicated capacity arrangements, and the broader enterprise infrastructure that has matured across 2023-2026.
Personal information handling remains a critical operational discipline. The original 2023 concern about inputting sensitive data into AI tools remains substantially relevant. Communications professionals continue to encounter operational security incidents where sensitive client information, internal strategy documents, or confidential material gets input into AI tools in ways that produce reputation and legal risk. The discipline of operational AI security has matured but the underlying risk pattern has not been substantially resolved.
The 2026 AI Integration Landscape
Contemporary communications operations integrate AI across six dimensions.
Research and synthesis. AI tools have substantially restructured research workflows — competitive intelligence gathering, executive briefing preparation, industry trend synthesis, and the broader research dimension of communications work. The discipline operates at substantially higher productivity than 2023-era workflows.
Drafting and ideation. AI tools support draft generation across press releases, executive communications, social content, talking points, and the broader written communications work. The integration produces substantial productivity gains when paired with human review and refinement.
Image and video generation. AI image generation (Midjourney, DALL-E, Stable Diffusion derivatives) and increasingly AI video generation (OpenAI's Sora, Google's Veo, the broader generative video category) have substantially restructured creative production workflows.
Translation and localization. AI translation has matured substantially across 2023-2026, producing professional-quality output in major language pairs and substantially reducing the cost and timeline of multi-market communications work.
Analysis and measurement. AI integration with analytics infrastructure has substantially improved sentiment analysis, brand monitoring, competitive intelligence, and the broader measurement dimension of communications work.
Agentic workflow automation. The 2024-2026 emergence of agentic AI (AI systems that can take multi-step actions with limited human supervision) has begun substantially restructuring how communications operations think about workflow automation. The category remains in early development but operates at meaningfully greater scale than the 2023 environment.
What Communications Professionals Should Actually Do in 2026
Five operational disciplines for AI integration in 2026.
Establish operational AI security protocols. Before integrating AI tools at any scale, communications operations should establish clear protocols for what information can be input into which tools, what data retention policies apply, and what review processes operate before AI outputs are published or distributed.
Maintain human review for all published content. AI-generated drafts should be reviewed and refined by human writers before publication. The discipline applies across press releases, executive communications, social content, talking points, and any other content that will be distributed externally.
Build measurement infrastructure for AI integration outcomes. Communications operations integrating AI at scale should measure productivity, quality, and risk outcomes. The discipline of measuring AI integration outcomes operates as foundational rather than optional.
Invest in AI Communications capability beyond AI tool usage. The category of AI Communications (how brands accumulate visibility and reputation inside AI engine answers) operates as a distinct discipline from AI tool integration (how communications teams use AI tools internally). Both disciplines require investment but they are not the same investment.
Maintain epistemic humility about AI capabilities. The category has evolved substantially across 2023-2026 and will continue to evolve substantially across 2026-2028. Communications operations that build flexible AI integration architectures consistently outperform operations that lock into specific tools or workflows.
