Search Engines vs AI Assistants

Search engines retrieve webpages you can evaluate yourself, rather than generating new text.

Search engines (e.g. Google) help you find information that already exists on the web.

• They scan billions of webpages and create an index — like a giant digital library.

• When you search, they retrieve relevant webpages from that index.

• You choose which sources to trust and read the content yourself.

Search engines like Google do use AI — but their main job is still retrieval, not generation. The output is always a list of links to real webpages you can visit yourself.

Tip: Looking for academic sources? Learn more in our Search and Discovery Tools guide.

AI assistants generate text rather than retrieve sources. Some now include web search, but you must still verify any claims they produce — and check any links or citations they provide.

• AI assistants, like Copilot Chat (Abertay’s supported AI assistant), create text by predicting likely word patterns based on massive training data. This training data can include academic papers, but also less reliable sources like blogs or forums.

• Some AI tools, such as ChatGPT, Claude, or Perplexity, now also include built-in web search. They act a bit like a research agent, blending live web results with generated text.

Be cautious: results are filtered, coverage is limited, answers may vary each time, and you must always check the originals before relying on them.

Tip: For practical guidance, see AI for researching a topic.

Many tools now mix search and generation, so the boundaries between them are less clear. For example:

• Google AI Overviews and conversational search — summarise web content, break down questions, and suggest steps with linked sources.

• Copilot Chat — available to Abertay staff and students, combining secure web access with conversational generation. Even with this secure access, Copilot Chat cannot reach all of the Library’s paid databases, so you must still use LibrarySearch and subject databases for your assessments.

• Perplexity — searches the web in real time while generating responses with citations. Popular for exploration, but not an academic search engine — it cannot access paywalled databases or provide citable sources.

• Academic platforms — e.g. Statista Research AI, which summarises licensed data alongside traditional results.

AI assistants act a bit like a research agent — doing some of the searching and filtering on your behalf. This means you don’t see the full picture and must be cautious:

• Filtered results — the AI chooses what to show you.

• Limited coverage — they rarely reach academic databases or paid-for Library resources.

• Check the originals — always go back to the source before you rely on it.

The key question: Can you trace and check the specific claims?

• With a search engine, you click through to evaluate the original source.

• With an AI assistant (even a hybrid one), you’re relying on the tool’s interpretation.

Tip: For reliable, citable material, always switch to academic tools like LibrarySearch, Google Scholar, or subject databases.

These issues arise because search engines and AI assistants rely on very different types of sources.

The difference between search engines and AI assistants is not just how they work, but what sources they use:

• Search engines retrieve indexed webpages, pointing you directly to sites you can evaluate.

• AI assistants generate answers from:

  • Training data (books, articles, websites, and datasets ingested during development).

  • Live web content accessed through built-in search features.

Why it matters:

• Search results are transparent lists of links you can check.

• AI outputs are blended, conversational answers that may be harder to trace.

Two further issues also affect whether AI outputs can be relied on for academic work: reproducibility and access to academic sources.

Reproducibility is an issue: the same AI query may not always give the same answer. This makes it harder to verify claims in academic work.

• Search engines are reproducible: if you run the same search, you’ll normally get a similar set of results (apart from minor ranking changes).

• AI assistants are variable: the same prompt can produce different outputs each time, because they generate text probabilistically.

Why it matters: In academic work, you need to show that your sources can be traced and checked by others. If an AI output cannot be reproduced, it cannot be used as reliable evidence.

Tip: Always go back to original articles, books, or data — don’t rely on an AI’s answer as your “source.”

This is especially important for postgraduate and funded research, where reproducibility is a core requirement.

Much academic and professional literature remains behind paywalls. These tools mainly rely on open-access research and freely available web content. Outputs often blend academic and non-academic material, without transparent, verifiable references.

Why it matters: If you rely only on AI assistants, you may miss key research articles, professional standards, or datasets that are not freely available.

• Library subscriptions: The University Library pays for access to specialist databases, journals, and ebooks that AI tools cannot normally reach.

Tip: Use AI for exploration, but switch to LibrarySearch and subject databases to make sure you include high-quality, peer-reviewed, paywalled sources in your work.

To access paywalled academic and professional literature, always use LibrarySearch and subject databases provided by the University Library.

AI can support exploration and brainstorming, but it cannot replace proper research. For academic work you must use verifiable, citable sources from LibrarySearch and subject databases. Always follow your assessment brief for rules on AI use.