ChatGPT citations after GPT-5.5: changes and impact on business visibility

ChatGPT Citations Have Changed: What SISTRIX Data Reveals

In May 2025, SISTRIX published an analysis that shook the AI search world. Following the rollout of GPT-5.5, citation patterns in ChatGPT responses underwent substantial modifications. Some domains saw their mentions drop by 40%, while others gained visibility with no apparent effort.

For SME and mid-market company leaders, this evolution is far from trivial. ChatGPT now generates over 100 million daily conversations. When a prospect asks "what management software for an industrial SME" or "best accounting firm in Lyon," the AI's response directly influences their decision. Being cited or not in these responses becomes a commercial imperative.

This article decodes the observed changes, identifies new criteria that favor citations, and proposes concrete actions to maintain or improve your visibility in AI engine responses.

What Has Changed in ChatGPT Citations Since GPT-5.5

Fewer Citations, But Higher Quality

SISTRIX data shows a 23% reduction in total citations within ChatGPT responses. However, this decline masks a deeper transformation. GPT-5.5 no longer cites sources just to "look credible." Every source mentioned now meets enhanced relevance criteria.

In concrete terms, generic citations have disappeared. An article that vaguely listed "10 digital marketing trends" without original insights is no longer referenced. Conversely, a detailed case study on a 34% conversion rate increase for a French B2B company gains visibility.

Redistribution of Cited Sources

Before GPT-5.5, major media and high-authority sites captured most citations. SISTRIX analysis reveals notable rebalancing:

• Traditional authority sites: 31% decrease in citations
• Specialized niche sources: 47% increase in mentions
• Content with proprietary data: 52% increase
• Product and service pages: 28% growth

This redistribution benefits companies that produce expert content in their business domain. A mid-market company specializing in precision machining, with a detailed technical blog, can now be cited as frequently as a generalist media outlet.

New Privileged Formats

GPT-5.5 treats content formats differently. Comparison tables, structured lists with quantified data, and precise definitions generate more citations. Long narrative paragraphs, even when well-written, are less frequently referenced.

At AISOS, we observe that pages structuring information into autonomous blocks, each answering a specific question, receive 3 times more citations than traditional linear content.

Why GPT-5.5 Modified Its Citation Criteria

Response to Hallucination Issues

OpenAI integrated enhanced factual verification mechanisms into GPT-5.5. The model now prioritizes sources it can cross-reference with other reliable information. This approach reduces hallucinations but also modifies cited content selection.

An isolated claim, even on a reputable site, will be less cited than information corroborated by multiple sources or supported by verifiable data. Companies publishing studies, benchmarks, or analyses with explicit methodology benefit from this evolution.

Better Understanding of User Intent

GPT-5.5 better distinguishes informational from transactional queries. For a question like "how to choose an ERP for SME," the model cites methodological guides. For "SME ERP pricing," it prioritizes commercial pages with pricing grids.

This sophistication means adapting your content to precise intents rather than targeting generic keywords. A page mixing advice and promotion will be less cited than a pure advice page or an assumed commercial page.

Contextual Freshness Consideration

Publication date weighs differently depending on the topic. For evolving themes like regulation, taxation, or technology, GPT-5.5 favors recent content. For stable subjects like proven management methods, age can be an asset if it demonstrates established authority.

New Criteria That Favor Citations

Named Entities and Verifiable Information

GPT-5.5 places greater importance on explicit entities. Mentioning "according to a study" no longer suffices. You must specify "according to the INSEE study from March 2025 on French SMEs." This precision allows the model to verify information and increases citation probability.

Apply this principle to your content:

• Name your sources with dates and authors
• Quantify your statements with precise figures
• Geographically locate your data when relevant
• Identify mentioned companies and individuals

Demonstrable Author Expertise

Content signed by identifiable experts is cited more than anonymous content or content attributed to generic brands. GPT-5.5 appears to cross-reference author information with other sources to assess legitimacy.

For an SME, this means highlighting internal leaders and experts. An article on industrial digital transformation signed by your technical director, with their biography and achievements, will be better treated than impersonal corporate content.

Clear Semantic Structure

HTML markup and logical content structure influence citations. GPT-5.5 more easily extracts information from pages that correctly use heading tags, lists, and tables.

Structural elements that improve citability:

• H2/H3 headings formulated as questions or direct statements
• Introduction paragraphs that summarize each section's content
• Bullet lists for enumerations and comparisons
• Tables for quantified and comparative data
• Explicit definitions for technical terms

Domain Thematic Coherence

GPT-5.5 evaluates a source's relevance relative to its apparent domain expertise. A cybersecurity-specialized site will be more cited on this topic than a generalist site, even if the latter has more traffic.

This evolution favors SMEs and mid-market companies that clearly position themselves on their business expertise. Publish regularly on your competency subjects rather than dispersing content across peripheral themes.

Adaptation Strategies to Maintain Your AI Visibility

Audit Your Existing Content

Start by identifying your pages currently cited by ChatGPT and those that have lost visibility. AISOS audits reveal that 60% of companies don't know which pages are referenced by generative AIs.

Recommended audit methodology:

• Test your main business queries on ChatGPT with search mode activated
• Note cited sources for each relevant query
• Identify competitors appearing in your place
• Analyze format and content differences between your pages and cited ones

Restructure Priority Content

Focus efforts on the 20% of pages generating 80% of your commercial value. For each priority page:

• Add explicit definition sections for key terms
• Integrate recent quantified data with sources
• Create section summaries at the beginning (TL;DR format)
• Structure information into autonomous, citable blocks

Produce High-Citability Content

Formats generating the most citations after GPT-5.5:

• Sector studies with your company's proprietary data
• Detailed methodological guides on business processes
• Factual comparisons between solutions or approaches
• Specialized glossaries with expert definitions
• Trend analyses with quantified projections

Concrete example: instead of an article "Benefits of Industrial Automation," produce "Industrial Automation in France: 2025 Study of 150 SMEs, Average ROI and Success Factors."

Strengthen Expertise Signals

Work on elements establishing your credibility:

• Complete author pages for each content contributor
• Press mentions and external citations to your site
• Visible and verifiable certifications and accreditations
• Detailed client cases with quantified results
• Publications in specialized media or journals

Measure and Adjust Your AI Visibility

Tracking Indicators

Measuring visibility in generative AIs remains challenging. Here are actionable metrics:

• Citation rate: appearance frequency on your target queries (weekly manual testing)
• Citation position: rank of your mention in source lists
• Referral traffic: visitors from ChatGPT links (identifiable in analytics)
• AI voice share: ratio of your citations vs competitors on a query panel

Adjustment Frequency

AI models evolve rapidly. GPT-5.5 will be followed by other versions, each with adjustments. Plan quarterly GEO strategy reviews to integrate observed evolutions.

Between these reviews, maintain vigilance on analysis tool publications like SISTRIX, and regularly test your priority queries to detect citation changes.

What These Changes Mean for Your Digital Strategy

The evolution of ChatGPT citations with GPT-5.5 marks a turning point for AI search optimization. Tactics that worked in 2024, focused on volume and domain authority inherited from traditional SEO, are losing effectiveness. The new paradigm rewards demonstrated expertise, proprietary data, and structure optimized for LLM extraction.

For SMEs and mid-market companies, this evolution represents an opportunity. You don't need a large corporation's communication budget to be cited by ChatGPT. You need content demonstrating your business expertise with concrete data and adapted structure.

Companies integrating these principles now gain significant advantage. Those who wait will see their AI visibility erode in favor of more responsive competitors.

Start by auditing your current visibility on priority business queries. Identify gaps with cited sources. Restructure key content according to GPT-5.5 favored criteria. This methodical approach will help you maintain and develop your presence in AI engine responses, where an increasing share of B2B discovery occurs.