The Complete Guide to Pre-Publication Paper Discoverability

Why Google Search, not just Google Scholar

For two decades, academic SEO meant one thing: optimise for Google Scholar. That advice is now incomplete. The 2025 data shows that the dominant surface where readers encounter your paper, including other researchers, clinicians, journalists, and AI retrieval tools, is Google Search, not Google Scholar. This chapter explains why, and what changes in how you write your title and abstract as a result.

The 2025 shift in one paragraph

Google rolled out AI Overviews as the default answer surface for most queries in 2024, and by late 2025 the effect on organic traffic was measurable and severe. Organic click-through rate dropped roughly 61% on queries where an AI Overview appears, according to Ahrefs (2026). Global publisher traffic from Google fell by about a third in the year to November 2025 (Press Gazette, 2026). Roughly 58% of clicks that once went to websites now resolve inside the AI summary itself without the user ever clicking through (PPC Land, 2025). If your paper is not formatted for AI-Overview extraction, it is no longer enough for it to rank on page one, page one no longer guarantees a click.

Ranking on Google does not mean being cited by Google's AI

The most consequential finding for academic authors is that citations in Google AI Overviews no longer track the top-ranked pages. Early in the rollout, roughly 75–76% of AI Overview citations came from the top-10 organic results. By February 2026, that overlap had dropped to between 17% and 38% (ALM Corp, 2026). In other words, being visible to a human searcher on Google is no longer the same thing as being cited when that same searcher asks an AI the same question. Different ranking, different winners, different optimisation targets.

AI retrieval tools are a third surface, with their own indexes

ChatGPT, Perplexity, and Claude do not use Google Search or Google Scholar as their retrieval backbones. ChatGPT uses Bing's real-time index, Perplexity continuously crawls the open web, and Claude mixes training data with its own search tool. Across all three, only about 12% of AI citations overlap with Google's top 10 organic results. Each platform has its own bias: Perplexity leans heavily on Reddit and real-time sources, ChatGPT favours Wikipedia and consensus documents, and Claude rewards structured, bullet-pointed content. A paper that is retrievable by one is not automatically retrievable by the others, and a paper that ranks on Google is not automatically cited by any of them.

The new hierarchy of surfaces

As a practical matter, a 2026 research paper needs to be discoverable on four distinct surfaces, which I will reference throughout the rest of this guide:

1. Google Search + AI Overviews. The largest surface by raw traffic. Your paper's title tag, meta description, and first abstract paragraph are what Google extracts for the result card and for AI Overview citations. Title truncation on Google Search is pixel-based, not character-based, with a practical limit around 50–60 characters on desktop (roughly 600 pixels) and shorter on mobile (Search Engine Land, 2025). Anything past that is either hidden or rewritten by Google itself, Google now rewrites around 60% of title tags at display time.
2. AI retrieval tools. ChatGPT, Perplexity, Claude, Elicit, Consensus, Semantic Scholar. These tools surface papers in answer to natural-language questions from millions of readers who never open Google Scholar. They extract claims almost exclusively from titles and abstracts. If your key finding is buried in your results section, an AI tool will cite whoever stated it more plainly in their abstract, even if your evidence is stronger.
3. Google Scholar. Still the primary surface for researchers doing systematic literature review and for citation tracking through its profile and "cited by" graph. Scholar has not had a meaningful algorithm update in years; the rules in this guide that are specifically about Scholar (the autocomplete test, profile setup, author-name normalisation) still hold. But Scholar is one of four surfaces now, not the only one.
4. PubMed and domain indexes. For biomedical papers, PubMed is still the gateway for clinicians and for literature-review software. For NIH-funded preprints, bioRxiv and medRxiv submissions are now surfaced in PubMed and PMC directly (NIH Preprint Pilot).

What this means for how you write your title and abstract

The short version is that a title optimised for Google Scholar alone is no longer sufficient. Four changes follow from the data above, and the rest of this guide applies them:

• Your title's first 60 characters must carry the finding. Google Search's pixel-based truncation will cut anything past that on the SERP, and AI Overviews and AI retrieval tools weight the first 60 characters disproportionately when extracting "what this paper shows".
• Your first abstract sentence must state the finding as a quotable claim. AI retrieval tools quote sentences; they do not paraphrase. Lead with the result, not the problem statement.
• Your abstract must contain explicit quantitative results. AI tools cite papers that state their numbers; they avoid papers that only gesture at them. "Improved activity >16-fold" belongs in both the title and the first two sentences of the abstract.
• Your metadata must work for all four surfaces. Most of the rules in this guide overlap across Google Search, Scholar, and AI tools, but where they diverge (for example, Scholar autocomplete versus Google Trends, or PubMed MeSH versus open-web keywords), I say so explicitly.

Title

The title is the single most important element for discoverability across every surface. On Google Search, your title is the blue-link text Google extracts for the result card and the snippet it cites in AI Overviews, and Google Search truncates it at roughly 600 pixels on desktop (about 50–60 characters) before rewriting or cutting what's past that (Search Engine Land, 2025). On AI retrieval tools, the title is the primary signal the system uses to decide whether your paper answers the user's question. On Google Scholar, the title carries far more ranking weight than any other part of the paper (Beel & Gipp, IEEE, 2009). An analysis of 140,000 papers (Letchford et al., 2015) confirmed that shorter titles receive significantly more citations. Most academic titles waste their most valuable space on filler words that nobody searches for.

What to do

1. Front-load the method, disease, or key finding in the first 60 characters. Google Search truncates titles around 600 pixels on desktop (roughly 50–60 characters), AI retrieval tools weight the opening of your title disproportionately, and Google Scholar gives stronger ranking weight to earlier terms. If your key terms are buried past character 60, most readers, human or machine, will never see them.
2. Use the exact terms people search for. Test your candidate title in three places: (1) regular Google's autocomplete, (2) an AI retrieval tool like Perplexity or ChatGPT typed as a natural-language question, and (3) Google Scholar's autocomplete. If none of the three suggest anything close to your title or return your paper, the terms are wrong.
3. Drop filler words. Remove "Novel", "A comprehensive study of", "Towards an understanding of", "Characterisation of", and "Insights into". These consume valuable title space and add no search value.
4. Include the specific technique or platform name. Researchers search for methods. "CRISPR base editing" is searchable. "Gene editing approach" is not.
5. Name the disease, organism, or tissue. "Pancreatic ductal adenocarcinoma" outperforms "cancer" for ranking because it matches what specialists search for.

The "after" title puts the method (engineered antimicrobial peptides), the pathogen (carbapenem-resistant K. pneumoniae), and the model system (murine sepsis) in the first 90 characters. A researcher searching for any of those terms will find this paper.

Abstract

Your abstract is the text Google Search extracts for the snippet that appears under your title on the results page, the block that AI retrieval tools quote when they summarise your paper for a natural-language query, and the source Google Scholar uses for its meta-description. On every surface, readers decide whether to click (or whether an AI cites you) based on the first two sentences. AI retrieval tools parse structured abstracts more reliably than unstructured prose, and they extract claims almost exclusively from the opening block, if your finding is buried in the implications paragraph, it effectively does not exist for ChatGPT, Perplexity, or Claude.

What to do

1. Front-load the first two sentences with your key finding and method. Do not open with "Background: Cancer is a leading cause of death worldwide." That wastes your most valuable SEO real estate on a statement no one searches for.
2. Use a structured abstract (Background, Methods, Results, Conclusions) even if the journal doesn't require it. Structured abstracts are parsed more reliably by Google Search, AI retrieval tools, Google Scholar, and PubMed, and AI tools in particular prefer bullet-pointed or clearly-labelled content when choosing which paper to cite (Discovered Labs, 2025).
3. Repeat your title's key terms in the abstract. Keyword co-occurrence between title and abstract is a ranking signal on Google Search, a retrieval signal for AI tools (they use title + first-paragraph cohesion to decide which paper answers a query), and a core ranking signal on Google Scholar. If your title says "CRISPR base editing", your abstract should include that exact phrase in its first two sentences.
4. Include quantitative results. "Reduced tumour volume by 68% (p<0.001)" is more informative and more likely to be cited by AI than "significantly reduced tumour volume."
5. End with a clear conclusion sentence that restates the key finding in slightly different terms. This gives search engines a second chance to match your paper to relevant queries.

Keywords

Journal keywords are indexed by Google Scholar, PubMed, and most database systems. They're also used by AI search engines to categorise your paper. Backlinko's analysis of 306 million keywords found that 91.8% of all search queries are long-tail phrases, specific, multi-word terms. Most researchers choose keywords casually, using broad terms instead of the specific phrases people actually search for.

What to do

1. Research actual search volumes. Use regular Google's autocomplete, Google Trends (filtered to "Science" category), PubMed's MeSH term browser, and Google Scholar's autocomplete to find the terms with highest search activity in your subfield. Cross-check against a natural-language query typed into an AI retrieval tool, if Perplexity or ChatGPT surfaces competing papers on your topic, note the phrases they actually quote in their answer.
2. Mix specificity levels. Include 2–3 highly specific terms (e.g. "carbapenem-resistant Klebsiella pneumoniae") and 2–3 broader terms (e.g. "antimicrobial resistance", "peptide therapeutics"). This captures both specialist and general searches.
3. Don't repeat words already in your title. Keywords should expand your paper's search footprint, not duplicate it. If your title says "CRISPR base editing", your keywords should cover related terms like "adenine base editor", "ABE8e", "haemoglobin disorders".
4. Use MeSH terms where possible. PubMed indexes papers against MeSH (Medical Subject Headings). Using exact MeSH terms as keywords improves PubMed discoverability.
5. Include synonyms and alternative spellings. If US researchers search "tumor" and UK researchers search "tumour", include both.

Introduction

The introduction is fully indexed by Google Search (via the publisher's HTML page), Google Scholar, and PubMed Central when the paper is deposited there. It is also the section AI retrieval tools draw from most heavily when they need to go beyond the abstract to summarise what a paper is about, Perplexity in particular will quote from your introduction when answering a reader's question (Discovered Labs, 2025). An introduction that buries its key terms under layers of generic context is harder for both human readers and retrieval systems to classify.

What to do

1. State the specific problem in the first paragraph. Don't spend three paragraphs on general background before mentioning what your paper actually addresses. Search engines weight earlier text more heavily.
2. Use your target keywords naturally within the first 300 words. This doesn't mean stuffing, it means ensuring the terms that appear in your title and keywords also appear in your introduction.
3. Name competing methods and approaches. If a researcher searches "advantages of base editing over HDR", your paper should mention both terms to rank for comparative queries.
4. Keep sentence length under 25 words on average. Long, convoluted sentences reduce readability scores. Lower readability correlates with fewer citations in most fields.
5. Minimise acronyms in the first paragraph. Search engines match plain text, not acronyms. Write "chimeric antigen receptor T cells (CAR-T)" in full before abbreviating.

Methods

The methods section is where researchers look for specific protocols, instruments, and techniques. It's heavily searched by people trying to replicate or adapt your approach. Including precise method names, software versions, and instrument models makes your paper findable for highly specific technical queries.

What to do

1. Name every instrument, software, and reagent by its full commercial name. "10x Genomics Chromium Single Cell 3' v3.1" is searchable. "Single-cell library preparation" is not.
2. Include software version numbers. Researchers search for "DESeq2 v1.38" or "Seurat v5 integration workflow". These are high-intent search queries.
3. Use subheadings that match how people search. "Cell culture and treatment" is clearer to both readers and search engines than "Experimental procedures".
4. Reference established protocols by name. If you followed the ENCODE pipeline or the Human Cell Atlas processing workflow, say so. These are searchable terms.
5. Describe statistical methods precisely. "Cox proportional hazards regression" ranks better than "survival analysis".

Results & Figures

Figure and table captions are independently indexed by Google Search, Google Images, and Google Scholar. They often appear in Google Search results as standalone snippets and in Google Images panels alongside the image itself. AI retrieval tools also quote figure captions when a reader asks a question about a specific result or chart. A well-written caption can drive traffic to your paper even when the main text doesn't rank for a particular query.

What to do

1. Write captions as self-contained descriptions. Each caption should be understandable without reading the main text. Include the key finding, method, and sample size.
2. Include searchable terms in every caption. "Fig. 3: Single-cell RNA sequencing reveals distinct CD8+ T cell exhaustion signatures in anti-PD-1 non-responders (n=24)" is searchable. "Fig. 3: Clustering analysis results" is not.
3. Use descriptive table titles. "Table 2: Baseline patient demographics" should become "Table 2: Baseline demographics of 312 patients with stage III NSCLC stratified by treatment arm".
4. Ensure all figures have alt text if you're submitting to a journal that supports it. Alt text is indexed by search engines.
5. Include units, sample sizes, and statistical significance in legends. These details make captions more informative and more likely to be cited by AI search.

Discussion

The discussion is where you connect your findings to the broader field. This is the section where AI search engines extract contextual information about what your paper means. Broader, higher-traffic search terms belong here, the discussion lets you rank for queries beyond your immediate niche.

What to do

1. Open with a clear restatement of the key finding. Don't assume the reader (or search engine) has read the results. State what you found before interpreting it.
2. Compare explicitly to named competing approaches. "Our lipid nanoparticle delivery system achieved 3.2-fold higher editing efficiency than the AAV-based approach reported by [Author et al.]" makes your paper findable for both "lipid nanoparticle delivery" and "AAV gene therapy comparison".
3. Use broader field terms here. If your methods section is about a specific assay, the discussion is where you mention "precision medicine", "immunotherapy", or "antimicrobial resistance", the high-volume search terms that connect your work to the bigger picture.
4. Address limitations honestly but concisely. Keep limitations to one paragraph. Excessive hedging reduces readability and doesn't improve search ranking.
5. End with clinical or translational implications if applicable. AI search engines heavily favour papers that state real-world relevance.

Conclusion

The conclusion is the last section indexed by search engines and is often extracted verbatim by AI platforms when generating summaries. A strong conclusion that mirrors the language of common search queries increases the chance your paper appears in AI-generated answers.

What to do

1. Write one sentence that could serve as a search result snippet. If someone searched for your topic, this sentence should answer their query directly.
2. Restate the key finding using slightly different terms from the abstract. This expands the range of queries your paper matches.
3. Include a forward-looking statement. "These findings support the development of peptide-based therapeutics for multidrug-resistant infections" is both a conclusion and a searchable statement about future directions.
4. Keep it to 3–5 sentences. Concise conclusions are more likely to be quoted in full by AI search engines.

References & Citation Strategy

Search engines use citation networks to determine a paper's authority and relevance. Citing well-indexed, highly cited papers creates backlinks in citation graphs that improve your paper's discoverability. This isn't about gaming the system, it's about ensuring your paper is connected to the right nodes in the citation network.

What to do

1. Cite the 3–5 landmark papers in your subfield. These are the papers that appear on page 1 of Google Search and Google Scholar for your target queries, and that AI retrieval tools repeatedly quote as the authoritative sources on your topic. Being in their citation network means your paper appears in Scholar's "cited by" lists that researchers browse and, increasingly, in the "related papers" panels that AI tools surface.
2. Cite recent papers (last 2–3 years). Google Scholar's "cited by" lists are sorted by relevance and recency, and AI retrieval tools disproportionately weight recent work when choosing which paper to cite for a current question. Citing recent work places you in active citation networks on both surfaces.
3. Cite papers from the journals you're targeting. Journal-level citation patterns influence how databases cluster related work.
4. Avoid excessive self-citation. Some search algorithms penalise papers with disproportionate self-citation rates.
5. Reference review articles in your field. Reviews are among the most-visited pages on Google Scholar and the most-cited sources when AI retrieval tools answer broad topical questions, they are the default landing page for any query like "what's known about X". Being in their "cited by" list, or being cited as a primary source inside a review an AI tool quotes, is valuable on both surfaces.

Lay Summary & Significance Statement

An increasing number of high-impact journals (Nature, PNAS, Lancet family, eLife) require or encourage lay summaries. These are indexed separately by search engines and are among the first content AI platforms extract when generating answers to general queries. A missing lay summary is a missed search opportunity.

What to do

1. Write at a Year 10 reading level (Flesch-Kincaid grade 10 or below). This isn't dumbing down, it's making your work accessible to the broadest possible audience, including clinicians, policymakers, and science journalists.
2. Lead with the real-world problem. "Antibiotic-resistant infections kill 1.27 million people per year" is more compelling and searchable than "Antimicrobial resistance is a growing concern."
3. State what you did and what you found in plain terms. "We designed new proteins that kill drug-resistant bacteria in mice" is clear, searchable, and quotable.
4. Include one sentence on why it matters. "This approach could lead to new treatments for hospital-acquired infections that don't respond to existing antibiotics."
5. Keep it to 100–150 words. Short enough for AI to quote in full. Long enough to be meaningful.

Graphical Abstract

Graphical abstracts appear on journal websites, in Google Images results, and are often the primary visual when papers are shared on social media or LinkedIn. Many journals (Cell Press, Elsevier) display them prominently on article landing pages. They drive clicks from visual search and social sharing.

What to do

1. Include your key finding as text overlay on the image. When the graphical abstract appears in Google Images or on social media, the text is what communicates the content.
2. Use a clear, left-to-right visual flow. Problem → Approach → Key finding works well.
3. Keep it to 3–4 panels maximum. Overly complex graphical abstracts are skipped.
4. Include the method name and disease/organism. These terms become alt-text metadata on journal sites.
5. Use high contrast and large fonts. Graphical abstracts are often viewed as thumbnails. Text should be readable at 200px wide.

Author Profiles & Metadata

Your author profile is the strongest single identity signal across academic surfaces. Google Scholar uses its profile to link your papers together and calculate your citation-based ranking signal, this is still the most important single step for researcher-facing discoverability. ORCID then links all your publications across Google Scholar, PubMed, CrossRef, and publisher sites, regardless of name variations or institutional changes. And an institutional page with a consistent bio is what Google Search and AI retrieval tools use to decide "who is this author" when a reader searches your name directly.

What to do

1. Complete your Google Scholar profile. Add a photo, verified institutional email, affiliation, homepage link, and at least 5 research interest keywords. An incomplete profile tells the algorithm you're not a strong signal.
2. Use a consistent author name across all publications. Decide on one format (e.g. "Sarah J. Chen") and use it for every paper, preprint, and conference abstract. If you've published under variants, manually merge them in Google Scholar.
3. Claim your ORCID and link it everywhere. Add your ORCID to your Google Scholar profile, institutional page, journal author accounts, and preprint servers. This creates a unified identity across systems.
4. Update your institutional webpage. Google indexes institutional pages highly. Ensure yours lists your publications with links, and includes your research interests as plain text (not only in a PDF CV).
5. Set up Google Scholar alerts for your own name. This helps you catch misattributed papers or profile fragmentation early.

Preprint Strategy

Preprints on bioRxiv, medRxiv, and similar servers are indexed by Google Scholar immediately, often within days. Journal publication can take 6–12 months. Meanwhile, Ahrefs (2025) found that AI Overviews reduce organic clicks by 58%, making early indexing even more critical. A preprint gives your paper 6–12 months of additional search indexing, citation accumulation, and AI training data inclusion before the journal version appears.

What to do

1. Check your target journal's preprint policy. Most major journals (Nature, Cell, Science, PNAS, NEJM, Lancet, BMJ) accept papers previously posted as preprints. Some have specific requirements about disclosure.
2. Post to the right server. bioRxiv for biological sciences, medRxiv for clinical and health sciences, ChemRxiv for chemistry, SSRN for social sciences. Field-appropriate servers have higher traffic from your target audience.
3. Use the same optimised title and abstract as your journal submission. The preprint version is what gets indexed first. Make sure it's optimised.
4. Post before or simultaneously with journal submission. This maximises the indexing advantage.
5. Update the preprint with the DOI when the journal version is published. This links the two versions and consolidates citation metrics.

Acknowledgements & Funding

Acknowledgements and funding statements are indexed and searchable. Funding bodies track their grants through text mining of published papers. Correctly naming your funder, grant number, and consortium improves discoverability in funder databases, institutional repositories, and compliance-tracking systems.

What to do

1. Use the funder's official name exactly. "National Institutes of Health" not "NIH". "Wellcome Trust" not "Wellcome". Many funders use automated text mining to find publications from their grants, exact name matching matters.
2. Include the full grant number. Funding bodies and their search systems link publications to specific grants. Missing grant numbers mean your paper may not appear in the funder's publication index.
3. Name consortia and collaborative networks. If your work is part of the Human Cell Atlas, TCGA, or UK Biobank, say so. These are high-traffic search terms.
4. Acknowledge core facilities and biobanks by name. Researchers searching for work using a specific facility or resource will find your paper.
5. Include your data availability statement with repository names and accession numbers. "Data deposited in GEO under accession GSE123456" makes your paper findable via the data repository.