Scholar Forensics

Investigation type 06  ·  AI Answer-Engine Audit  ·  verdict: cited, misdescribed, or absent

When a researcher asks an AI which journal to cite, does yours appear?

Researchers now ask ChatGPT, Claude, and Perplexity which journals and papers to trust before they open a search engine. If the models don't name your journal, you lose readers and authors at the recommendation stage — invisibly, before a single click. This audit measures where your journal stands, and why.

Engines audited: ChatGPTClaudePerplexityGeminiGoogle AI Overviews
Definition

LLM visibility for a scholarly journal is the degree to which large language models cite the journal or its articles when a researcher asks which sources are authoritative in a field.

A journal has high LLM visibility when answer engines name it accurately and often; low visibility when the models cite competing journals or omit it entirely. It differs from Google Scholar indexing: indexing governs whether articles are findable in a search index; LLM visibility governs whether they are recommended inside a generated answer — frequently with no click ever reaching the journal.

The problem, stated plainly

Absence from AI answers is a discoverability failure that standard analytics cannot see.

A journal can be fully indexed in Google Scholar and still be invisible in ChatGPT, because the two systems select sources by different mechanisms. When an answer engine shapes a researcher's shortlist without sending a click, the loss never appears in your traffic reports. It looks like nothing is wrong — while prospective authors are quietly routed to the journals the models do cite. This audit makes that invisible loss measurable.

What the audit delivers  ·  six phases

  1. Field-specific prompt set

    30–50 real research prompts in your journal's subject area — not generic brand queries. The questions researchers actually ask: who wrote the seminal work on a topic, where to publish on it, which journals lead the field.

  2. Multi-engine run

    Every prompt run across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. For each: is the journal named, which articles are cited, and how accurately is it described. Citation behavior differs by engine, so all are tested.

  3. Competitor share of voice

    The same prompt set run against 3–5 competing journals. The result is a clear ranking of who dominates the answers in your field — and by how much.

  4. Citation-source diagnosis

    Where the models draw their answers from — which domains, databases, and pages they cite for your field — and why your site is or isn't among them.

  5. AI-readiness technical check

    Crawler access for the AI user-agents, full-text and metadata machine-readability, structured data, and llms.txt — the site-side signals that determine whether your content is even eligible to be cited.

  6. Written verdict and remediation plan

    A dated report: the finding, the evidence behind it, a stated confidence level, and a prioritized, DOI-safe plan for what to change. The prompt set ships with it, so you can re-run and verify yourself.

Questions publishers ask

How is this different from an ordinary AEO or SEO audit?
Most AI-visibility tools are built for commercial brands and local businesses, using generic prompts and brand-mention counts. This audit is built for scholarly journals: it reasons about DOIs, full-text crawlability, Scholar and Scopus mechanics, and how models treat academic sources as citations rather than as brands. The subject expertise is the difference, not the tooling.
Can you guarantee my journal will be cited after the fixes?
No, and any provider who guarantees a citation is misrepresenting how these systems work. Model output cannot be directly controlled. What the audit does is remove the reasons for invisibility — crawler blocks, unreadable metadata, missing structured data, weak third-party presence — so the journal becomes eligible to be cited. Absence of those fixes almost guarantees invisibility; their presence makes citation possible, not certain. This honesty is the point of a forensic method.
Is the measurement exact?
It is directional, not exact. Model answers vary between runs. A fixed prompt set measured against a dated baseline makes the movement reliable and repeatable, which is what matters for tracking change over time.
Does this replace Google Scholar or Scopus work?
No — it complements it. Scholar and Scopus govern index findability; LLM visibility governs answer-engine recommendation. A complete discoverability picture needs both, which is why this sits alongside the deindexation and migration investigations rather than replacing them.

Deliverable & engagement

Format
Written report (DOCX) + prompt-set table you can re-run
Baseline
Dated, so change is measurable on re-audit
One-time audit
Fixed scope · quoted per journal & field breadth
Optional watch
Quarterly re-run · visibility drifts as models update

Start an audit

Tell me your journal, its field, and the competitors you watch.

You'll get a first read on where you stand across the answer engines — with evidence, not a guess. Any platform: WordPress, Joomla, OJS, or custom.

Request an LLM visibility audit Paid diagnostic · written findings · confidence level stated · no guaranteed-citation claims