- Choose one reader, one field and the kind of post you will write.
- Make a source-to-claim record before you draft.
- Draft one clear explanation that shows its limits.
- Put every number in context: the denominator, the time frame and the comparison.
- Use visuals that explain real evidence and say what they are.
- Choose a platform that renders your equations and tables, then set up the site.
- Check the post against its sources, publish it and share it where the rules allow.
- Keep the archive current, log corrections and disclose any paid links.
How to Start a Science Blog in 2026: 8 Steps for Beginners
- 23 min read
- 10 sections
- 4 comparison tables
- 15 questions answered
The short answer
Want the site rather than the reading? Try Grigora free for 14 days
Contents10 sections

Read nextHow to Make a Lab Website: Pages, Platforms, Costs, Examples
Is a science blog worth starting in 2026?
Fewer outlets now explain science to the public. In a September 2026 article republished by Nieman Lab from The Conversation, Bruce Lewenstein and Chris Mooney counted 29 prominent popular science magazines started in the United States since the 19th century and found only 11 still running, a preliminary count. Scientific American’s new owners laid off 15 staff after buying it in late June 2026, and Popular Science had no digital magazine by the end of 2023. Careful independent explainers can fill part of that gap, but you will have to build your own readership. The same authors note that the growth in online science media “tends to favor quick takes and attitude”, which is why steps 2 to 4 below matter more than publishing fast.
1. Choose the reader, subject and kind of post
“Science” is too broad to guide a first draft. Choose a field and a reader’s problem: an interested beginner trying to understand a term, a student comparing methods, or a practitioner who needs a clear account of a particular result. Describe the background knowledge you assume and the new understanding the post should provide.
| Format | Useful starting material | Boundary |
|---|---|---|
| Concept explainer | Reliable subject references and one clearly defined question. | An analogy needs its limits stated; it is not a literal model of every detail. |
| Research summary | The full paper or report, relevant background and expert interpretation where needed. | A press release or abstract alone may leave out methods, uncertainty and limits. |
| News reaction or myth check | The original study behind the headline, plus the coverage you are responding to. | Link and read the original study, not just the coverage of it. |
| Data explainer | A published dataset or chart with its units, method notes and license. | Check the dataset license and units; material from third parties keeps its own terms. |
| Methods or research diary | Your own permitted work, records and explanation. | Do not invent experiments, access, affiliations, data or laboratory experience. |
| Interview or review | A real interview, or a tool or work you have evaluated yourself. | Keep attributed statements and observed features apart from your own interpretation. |
You can write as a learner or enthusiast without claiming specialist credentials. State your experience honestly, show your sources and get suitable review for claims beyond your expertise. A student explaining a method they are learning, or a teacher explaining a topic they teach, can write a useful post when readers can see where each claim comes from.
If you plan to blog about your own research, check with your PI or institution first, especially before writing about unpublished data, lab techniques, embargoed results or anything that could identify participants. A personal blog is not an exception to those rules. Assume supervisors, colleagues and future hiring panels will read it; Bitesize Bio’s advice on starting a science blog makes the same point about future employers. Guest posting is a lighter way to start: The Conversation US takes pitches from scholars and researchers affiliated with a college, university or research institution, and PhD students can write about the topic of their own research.
First science blog post ideas
If you are stuck on a first topic, pick one you can source well this week. Some starting points:
- What a viral headline’s study measured: link the paper and compare its sample and outcome with the headline.
- A term from this week’s news explained, such as a statistical or laboratory term readers keep meeting.
- How a common method works, for example PCR, carbon dating or a randomized controlled trial.
- A data explainer built on an Our World in Data chart, which its FAQ lets you reuse with credit; charts made by third parties keep their own license terms.
- A monthly roundup of a few papers in your field, each with its status and one limit.
- A myth check: trace a common claim back to its source and say what the evidence supports.
- A real interview with a researcher about one paper, with their answers attributed.
- A reading list of review articles for a beginner in your field.
- A local science story, such as what your water supplier’s published quality report measures.
- A “what we still do not know” post on an open question in your field.
2. Make a source-to-claim record
Before drafting, open the original source and record what it actually establishes. For a research story, read the methods, results and limitations as well as the abstract. Put the finding in the context of appropriate reviews, established references or guidelines; one new paper does not automatically replace the wider evidence. For a basic explainer, choose references suited to the claim rather than adding a paper only to look authoritative.

| Field | Question to answer |
|---|---|
| Source and version | Author, title, date, DOI or stable URL, version and the date you accessed it. Is it a paper, preprint, review, guideline, dataset or press release? |
| Question and design | What was asked, measured or modeled? What comparison and assumptions were used? |
| Population, material and context | Who or what was studied, with what sample, setting, conditions and time frame? |
| Result and uncertainty | What changed, in what units and by how much? Which uncertainty or limitations matter? |
| Interpretation | Which statements are the source’s findings, the authors’ interpretation, outside context or your own explanation? |
| Status, interests and permissions | Corrections or retractions, funding and conflicts of interest, the license for data and figures, and any publication approval you need. |
| Claim decision | Which exact sentence can the evidence support, which qualification does it need and who reviewed it? |
For health-related research, MedlinePlus’s guide to evaluating health information lists questions to ask of any source: who runs the site, who pays for it, how content is reviewed, where it comes from and when it was updated. Do not turn a laboratory or animal finding into a demonstrated benefit for people, or an association into a claim of cause, without a study design that supports it. Medical guidance that people might act on needs evidence and expertise suited to the decision.
Before publishing, check the publisher’s page for updates. Where available, Crossref’s Crossmark can show corrections, retractions and other changes, but its presence is not a guarantee of accuracy or complete coverage, and its absence does not prove a paper is unchanged. Also search the title or DOI in the Retraction Watch Database, which Crossref acquired in September 2023 and made publicly available. Retractions are not rare: a Nature analysis counted more than 10,000 retracted research papers in 2023, a record. Record the version you used.
Label a preprint’s status and version plainly. medRxiv’s FAQ says its preprints are not certified by peer review and should not be used to directly inform clinical decision-making, and it asks people reporting on them to stress that the work has yet to be evaluated by the medical community and that the information may be erroneous. Do not describe a preprint as peer reviewed, or imply that peer review makes every published conclusion final. Check for a later journal version and note any material differences.
3. Draft one clear explanation with visible limits
Give the reader the answer or main finding early, then explain the route to it. Define unfamiliar terms when they first matter. Use the detail needed for the intended audience; a public explainer and a specialist methods post can have different levels of technical depth. Keep the exact meaning when simplifying.
- Reader question: state the narrow question the article answers and its scope.
- Short answer: explain the central point with the essential qualification.
- How we know: describe the method or evidence in plain terms, with a source beside the claim.
- Explanation: use a worked example, labeled diagram or comparison that helps the reader.
- Limits: explain uncertainty, competing interpretations, assumptions and what cannot be concluded.
- Next step: link background reading and the sources, and say who wrote and reviewed the post, any interests and when it was updated.
Some publishers offer rules of thumb for readable research posts. De Gruyter’s advice on blogging about your research suggests sentences under 20 words, paragraphs of no more than 600 characters, titles of 10 to 12 words with the key terms near the start, and testing the draft on 2 or 3 people from your target audience, preferably non-experts. Treat these as suggestions rather than rules: a methods post for specialists can run longer. The free readability score checker gives a quick reading-level check on a draft before you share it.
Ask a reader from your intended audience which sentence or visual they could not follow. Ask a knowledgeable reviewer whether the explanation is still accurate. These are different checks, so record which one each person did. Do not add an expert-review badge or a quote without a real review or interview.
AI can help organize supplied notes or revise phrasing, subject to the rules for your work. Check every citation, quotation, number and factual statement against the real source. Do not generate research findings, expert endorsements, participant experiences or evidence images. Keep confidential data out of tools that are not approved for it.
4. Explain numbers without hiding the denominator
Give the comparison, denominator, units and time frame a reader needs. Distinguish a percentage change from a change in percentage points. Report uncertainty and important study limits when they affect the interpretation. A dramatic relative number can sound much larger than the underlying absolute difference, and a 1995 warning about contraceptive pills in the UK is a well-documented example.
| How the 1995 warning could be reported | Value |
|---|---|
| Relative change (the headline) | Risk “doubled”, a 100% increase |
| Thrombosis on older pills | 1 in 7,000 women |
| Thrombosis on third-generation pills | 2 in 7,000 women |
| Absolute increase | 1 extra case per 7,000 women |
| What followed | An estimated 13,000 additional abortions in England and Wales in the following year |

The same evidence supports both “the risk doubles” and “1 extra case in 7,000 women”, but only the second tells readers how many people are affected. The figures are as reported in Gerd Gigerenzer’s 2009 editorial in the Bulletin of the World Health Organization, and the framing had consequences: a 1999 review in Human Reproduction Update linked the scare to a substantial increase in unintended pregnancies, particularly among younger women. When you report a risk, give both numbers, the denominator, the time frame and who was compared. Gigerenzer’s editorial does not say what period the 1 in 7,000 covers, so a full report would take that from the original studies rather than guess it. This is a historical example of risk communication, not advice about contraception. Do not add precision or an uncertainty range that the source does not give.
5. Use visuals that explain the actual evidence
Choose a visual because it answers a question, shows a relationship or clarifies a process. Label whether it is a diagram, model, simulation, measured image or illustrative example. A stylized cell, planet or laboratory picture must not look like documentary evidence of the study being discussed. Do not invent a microscope image, plot or experimental result.
- For a data chart, name the real dataset or source, the variables, units, comparison and time frame.
- Check scales, baselines, missing values, aggregation and uncertainty; do not choose a crop or chart type that changes the meaning.
- For a process diagram or analogy, label the simplifications and say where the comparison stops working.
- Check reuse rights for each figure and every third-party element in it. The PMC copyright notice explains why free access does not remove reuse restrictions, including on third-party material, and our guide to using images from the internet covers licenses and credit.
- Give the required credit and a caption that says what the figure shows. Put the essential explanation or data in the text too, rather than relying on color or a small image alone.
The W3C complex-image guidance explains how a short and a longer text description can work together for informative charts and diagrams. Check each figure on a phone too: axis labels and source notes that are readable only at full desktop width need a different layout, such as a stacked phone version.
Also worth readingHow to Start a Technical Blog: Platforms, Costs and a First Post
6. Choose a platform and prepare the site
Before you choose, build one test post that contains the hardest things you plan to publish: citations, a table, a captioned figure and any equations or code. Check that they stay readable on a phone, in link previews and with a screen reader if you can. A science-themed template is optional; an editor that handles your material is not. The table compares six common routes.
| Platform | Equations | Cost model | Best fit | Watch for |
|---|---|---|---|---|
| WordPress.com | Math block and inline math, written in LaTeX | Free plan with 1 GB of storage; a custom domain and plugins need a paid plan | A classic blog on a hosted service | Check which features your plan includes before you build around them |
| Blogger | No built-in tool; add a MathJax script to the theme | Free Google service | A simple free blog | A custom domain runs on a subdomain such as www or blog; test the script on every post |
| Substack | LaTeX block, added in January 2023 | Free to publish; with paid subscriptions, Substack keeps 10% plus Stripe’s fees | A newsletter-first writer | Posts can go out as emails too, so check how equations and tables read in an inbox |
| Ghost | No equation block; add KaTeX with Code Injection (community guide) | Open-source software; Ghost(Pro) hosting is paid, from $15 a month billed yearly | A blog with a built-in newsletter | The cheapest Ghost(Pro) plan has no paid subscriptions |
| Grigora | No equation block; test a KaTeX or MathJax script in custom code | 14-day free trial with no card, then Starter at $15 a month, or $144 a year ($12 a month) | A blog inside a fuller personal, lab or outreach site drafted by AI | No reference manager and no paid subscriptions |
| Self-hosted WordPress | Plugins for LaTeX and citations | Free software; you pay for hosting and a domain | Full control over plugins and themes | You handle hosting, updates, security and backups |
Vendor terms were checked on 16 November 2026: WordPress.com LaTeX guide and plan features, Blogger custom domains, Substack costs and its January 2023 LaTeX announcement, Ghost Code Injection, the Ghost forum’s KaTeX thread, Ghost pricing and Grigora pricing. Plans change, so confirm them before you pay.
Equations, references and code
Equations are where platforms differ most. The Equations column above shows what each platform offers; where there is no built-in block, test the script with your real equations before you commit. Whichever you use, check a long equation on a phone and explain it in words for readers who skip the notation. For references, link each claim to the paper’s full https://doi.org/ address, name the journal or preprint server and the year, and say whether it is a preprint; a short list of key references at the end helps readers who want more. The DOI Foundation explains how DOIs work. If your posts lean on code, our guide to starting a technical blog covers code samples and reproducible tutorials.
A domain name should be readable and suit the subject, with conflicts and renewal costs checked before you buy it. The free domain name generator can suggest options to check. A custom domain is not a requirement on day one: a platform address can suit a first project. Do not use a name or logo that falsely implies an institution, journal or research group endorses the site.
Grigora suits a narrower case: a researcher, science communicator or educator who wants the blog inside a fuller site. Describe your research focus or outreach work and Grigora’s AI drafts the pages, such as About, research topics, a blog and a contact page, which you then change by prompt. The blog is a CMS collection, and you can add another collection for publications. Newsletter signup forms and campaigns are included (the trial allows 50 subscribers and one campaign a month), and visitor comments wait for your approval before they appear. If you already blog elsewhere, Settings > Migration imports posts, pages, categories, tags and authors from WordPress, Blogger, Ghost, Wix, Webflow and Duda. Paid plans include hosting, SSL, a CDN and connecting a domain you already own; Grigora does not sell domains. Starter costs $15 a month, or $144 a year ($12 a month), for one website after a 14-day free trial with no card. The limits matter for science writing: there is no equation block or reference manager, and there are no paid subscriptions, only one-time payments. For research pages beyond the blog, see our guide to making a lab website.
Add an accurate About page and byline with your relevant background, a way to contact you and report errors, and a note on how you update posts. The free About Us page generator can give you a first draft to edit. If you collect comments, analytics or email signups, explain how you use the data and meet the rules that apply to you. A privacy policy generator gives you a starting point, not an answer: a generic template should not claim practices the site does not follow. Finally, check account recovery and how you would back up or export your posts.
7. Publish and share an accurately checked first post
- Read the headline, description, body, captions and social preview together: do they all describe the same supported claim?
- Check every quotation, number, unit, link, DOI and version or status notice against the source.
- Confirm what expert or editorial review took place, and any institutional, participant or media permissions you need.
- Open the page on a phone and a desktop; test tables, any equations or code, source links, image descriptions and navigation.
- Visit the final public URL as a reader would and test the contact and correction route.
- Share a faithful preview in relevant communities, follow their rules and do not use a stronger headline than the evidence supports.
Search Engine Optimization (SEO) should make the subject clear and the page easy to use. Write a descriptive title and headings, a distinct summary and helpful related links; the free SERP preview tool shows how a title and description may look in search results. Google’s guidance on helpful, people-first content does not say that a long article, a science keyword or frequent publishing will lead to higher rankings. A correct, modest finding serves readers better than an unsupported “breakthrough”.
Share where your readers already are: your own newsletter list, subject communities whose rules allow links, and replies to people already discussing the paper you explain. A guest post, such as a pitch to The Conversation if you qualify, can introduce you to new readers. Invite questions about the explanation and the evidence, and moderate comments so they wait for approval before they appear, because contested topics can draw hostile replies. If your subject draws abuse, consider writing under a pseudonym; our guide to starting an anonymous blog covers what a pen name does and does not hide. When you correct a post, correct it where the original claim circulated too, including a misleading preview or chart. No channel guarantees readers, collaboration or income.
8. Keep the archive and commercial relationships honest
Regular posting can help you organize your work, but source checks and corrections need time too. Choose a cadence you can keep and maintain a list of ideas and sources. A short, well-supported explainer may serve readers better than an unnecessarily long post, and there is no required number of posts before launch. Revisit important source updates and unanswered reader questions.
Monetizing your blog is optional, and none of these routes promises income. Common options are paid subscriptions on Substack, where Substack keeps 10% of subscription revenue and Stripe charges its own fees; display ads; affiliate links with a disclosure right next to the link (the FTC’s Endorsement Guides FAQ says a “paid link” label next to an affiliate link should be adequate); commissioned writing or talks; and one-time paid resources such as revision guides. If you teach, Grigora Commerce can sell digital or physical products as one-time payments, paid into your own Stripe, PayPal, Paddle, Razorpay or Paystack account with no per-sale fee from Grigora. Keep financial relationships separate from the strength of the evidence, follow the FTC’s Disclosures 101 guidance and any other rules that apply to you, and never promote a scientific or health claim beyond its evidence.
Keep an editorial log with the changed passage, the reason, the sources and the date. Make substantive corrections visible to readers and recheck affected figures, downloads and previews. Track useful questions and comprehension feedback alongside traffic. Page views or sales do not show that a claim is true or that readers learned something.
Frequently asked questions
Practical writing on building a site, growing an audience and sending email people open. No more than one a week.

Karan Singh Bhakuni
120 articles on the Grigora blog
More on Start a Blog

How to Make a Lab Website: Pages, Platforms, Costs, Examples
How to build a research lab website: the pages to include, six routes compared by yearly cost, four real lab sites worth studying, publication, privacy and accessibility checks, and how long a build…
24 Aug 202320 min read

How to Start a Technical Blog: Platforms, Costs and a First Post
Choose a narrow reader, compare DEV, Hashnode, GitHub Pages, WordPress and hosted-site costs (checked November 2026), cross-post with canonical links and publish a tested first post.
12 Oct 202321 min read

How to Start a History Blog: 8 Steps, Ideas and Sources
A practical way to start a history blog: choose a focus you can research, study real examples, set up the right pages, find free archives, check image rights and publish your first posts.
22 Sept 202324 min read