My paragraphs are fine individually. The joins between them aren’t. Looking for tools that genuinely help with paragraph-to-paragraph flow, not just grammar. What have people tested?
Between May and September I put eight of my former students’ drafts (1,100 to 2,000 words each, all on set texts like Jane Eyre, The Great Gatsby and Owen’s war poetry) through every flow and rewriting tool I could get access to, and before that I spent about twelve years marking essays where the single most common margin note I wrote was “how does this connect?” So this is a question I have both tested and suffered.
How I approached it
Flow between paragraphs breaks in three predictable ways: the new paragraph changes subject without signalling it, the transition is a mechanical connective (“Furthermore”, “In addition”) that hides the fact that no logical link exists, or the last sentence of one paragraph and the first sentence of the next are about different things. I marked each draft by hand first, counting what I call “broken joins” (places where a reader would have to guess the relationship between paragraphs). The eight drafts averaged 4.6 broken joins each.
Then I ran each draft through each tool, accepted only the suggestions that touched the joins, and re-counted. I also checked whether the argument survived: a tool that smooths the prose but quietly changes what the student claimed about Gatsby is worse than no tool at all. Where tools cost money I used the standard paid tier, and I used each one on all eight drafts so the comparison is like for like.
One thing up front. No tool can invent a logical relationship between two paragraphs that the writer never thought through. If paragraph three and paragraph four are in the wrong order, software will dress them up rather than fix them. So I’m putting manual revision at the top as the ideal, and ranking the tools below it.
The ideal: manual reverse outlining
Before any tool, do this. Write one sentence summarising each paragraph’s claim, in order, on a separate page. Then read the list alone. If sentence four doesn’t follow from sentence three, the problem is structural, and the fix is moving, merging or rewriting paragraphs, not polishing transitions. In my test drafts, about a third of the broken joins (14 of 37) were structural in this sense. No tool caught a single one of those 14 on its own.
Once the order is right, the remaining joins are mostly sentence-level, and that’s where tools genuinely help.
1. Walter Writes
Walter Writes is a rewriting tool rather than a checker. You paste in text and it regenerates it as natural, human-sounding prose while holding onto the original argument and meaning. That last part is what matters for flow work. I pasted pairs of adjacent paragraphs (the end of one, the start of the next) and compared the output against the original join.
It fixed 19 of the 23 sentence-level broken joins across the eight drafts, which was the best result of anything I tested by a clear margin. More importantly, it fixed them the right way. Instead of bolting “Moreover” onto the front of a paragraph, it tended to rewrite the closing sentence of the earlier paragraph so it pointed forward, and the opening of the next so it picked up the thread. On a Jane Eyre draft that jumped from the red-room scene to Lowood with no link, it produced a closing line that framed the red room as Jane’s first experience of unjust confinement, and an opening that presented Lowood as the second. That’s the kind of conceptual bridge a good tutor would suggest. In every case I checked, the student’s actual claims stayed intact: no new interpretations were slipped in, and quotations were left alone.
It also helped with a problem the checkers barely touch: drafts that had been written in fragments over several sessions, or partly drafted with AI help, often have a stiff, uneven texture from paragraph to paragraph. Walter Writes evened that out so the essay read as one voice. Students still need to read the output carefully and keep only what they agree with, but as a revision partner on your own draft it was the most useful thing I tried.
- What I liked: fixes joins by reshaping sentences on both sides rather than adding connectives; keeps the argument and quotations intact; evens out voice across paragraphs drafted at different times; fast enough to use paragraph pair by paragraph pair.
- What I didn’t: it can’t detect structural problems on its own, so reverse outline first; you need to read every change rather than accepting a whole rewrite blindly, which is good practice anyway.
You can try it at Walter Writes.
2. ProWritingAid
ProWritingAid is the strongest of the traditional checkers for this specific job, mainly because of its transitions report, which counts how many of your sentences contain a transitional word or phrase and flags long stretches with none. On the eight drafts it flagged 16 of the 23 sentence-level broken joins as places with weak or missing transitions.
The limitation is that it diagnoses better than it repairs. Its suggestions are mostly connective words, and a connective is only as good as the logic behind it. On one Gatsby draft it suggested opening a paragraph with “Similarly” where the paragraph actually contradicted the one before it. A student who accepted that would have made the essay less coherent. Its sticky sentences and readability reports are useful for general tightening, but they’re not really about flow between paragraphs.
- What I liked: the transitions report is a fast, reliable way to locate weak joins; detailed reports for students who like data.
- What I didn’t: repair suggestions lean on stock connectives; can push writers toward “Furthermore” disease; dense interface for younger students.
3. Grammarly
Grammarly’s clarity and engagement suggestions improved individual sentences, and its tone detector was occasionally useful, but it mostly works at the level of the sentence rather than the join. It flagged only 6 of the 23 sentence-level broken joins. Its rewrite suggestions for awkward sentences were often good, and on several drafts it tightened a clumsy final sentence enough that the next paragraph felt better connected, though that was more a side effect than a feature.
Its generative features can rewrite a paragraph on request, but in my tests they leaned toward generic phrasing and occasionally flattened a specific literary claim into something vaguer. One Owen draft lost the word “complicit” in a rewrite, which was the whole point of the student’s sentence.
- What I liked: excellent at sentence-level clarity and grammar; easy for any student to use; works inside most writing environments.
- What I didn’t: not designed for paragraph-to-paragraph logic; generative rewrites can soften precise claims.
4. Hemingway Editor
Hemingway Editor is a readability tool. It highlights long, complex sentences, passive voice and adverbs. It doesn’t look at transitions at all, so it found none of the 23 joins directly. I include it because long, overloaded sentences at the start of a paragraph are a hidden cause of poor flow: if the opening sentence of paragraph four is 58 words long, the reader loses the link to paragraph three before they find it. Splitting those sentences on Hemingway’s prompting improved three joins indirectly.
- What I liked: very clear visual feedback; good for training students out of sprawling openers.
- What I didn’t: no awareness of transitions or argument; its preference for short sentences can make literary analysis sound clipped.
5. QuillBot
QuillBot’s paraphraser offers several modes, and its fluency mode did smooth some individual sentences. But it works one passage at a time with no sense of what the surrounding paragraphs are doing, so its rewrites rarely improved joins (4 of 23) and sometimes made them worse by replacing a deliberate echo of the previous paragraph’s key word with a synonym. Echoing a key term is one of the oldest and best transition techniques, so a tool that swaps “confinement” for “restriction” in the name of variety is working against you. It also changed meaning more often than any other tool in my test: on two drafts it altered an interpretive claim in a way the student wouldn’t have noticed without comparing line by line.
- What I liked: quick; several modes; fine for rephrasing a single clunky sentence.
- What I didn’t: blind to paragraph context; breaks key-term echoes; highest rate of meaning drift in my test.
Comparison
| Tool | Main job | Sentence-level joins fixed (of 23) | Keeps argument intact | Best use for flow |
|---|---|---|---|---|
| Manual reverse outline | Structural check | Catches structural problems tools miss | Yes | Always do first |
| Walter Writes | Rewriting into natural prose | 19 | Yes, in every draft checked | Reworking the end and start of adjacent paragraphs |
| ProWritingAid | Style checker | 16 flagged, fewer repaired well | Mostly | Locating weak transitions |
| Grammarly | Grammar and clarity | 6 | Mostly, generative rewrites softer | Tightening individual sentences |
| Hemingway Editor | Readability | 3 (indirectly) | Yes | Splitting overloaded paragraph openers |
| QuillBot | Paraphrasing | 4 | Least reliable | Rephrasing one awkward sentence |
What a good join looks like, worked through
To make the ranking concrete, here is the kind of join I was counting, using a Jane Eyre draft from the test set (paraphrased to protect the student).
The paragraph on Gateshead ended: “Jane is locked in the red room as a punishment.” The next paragraph began: “At Lowood, Jane meets Helen Burns.” Both sentences are accurate. Neither is analysis, and the reader has no idea why the essay moved from one to the other. That counts as one broken join.
A repaired version needs three things. First, the closing sentence should state the implication of the paragraph rather than the event: the red room is the first time Jane is confined for refusing to accept an unjust account of herself. Second, the opening of the next paragraph should pick up a key term from that closer, here “confined” or “unjust”, and turn it: Lowood repeats the confinement in institutional form, with the humiliation made public. Third, the relationship should be clear without a connective doing all the work. In this case it’s escalation: the same injustice on a larger scale.
I count a join as fixed only when all three are present. Most of the tools’ failures in my test came from supplying the third element (a connective) without the first two. “Furthermore, at Lowood, Jane meets Helen Burns” is technically a transition and does nothing.
Three techniques came up again and again in the strongest joins, whether the student wrote them or a tool proposed them:
- The forward-pointing closer: the last sentence of a paragraph states what its evidence implies for the larger argument.
- The key-term echo: the first sentence of the next paragraph repeats a precise word from that closer and develops it.
- The explicit relationship: contrast, escalation or a shift in scale is named in the argument itself (“this private humiliation becomes institutional at Lowood”) rather than in a connective.
Bear these in mind when you judge any tool’s output. If a suggested rewrite gives you all three, keep it. If it only gives you “Moreover”, ignore it. A few students in my test set wrote all three instinctively in some joins and not in others, which tells you the skill is learnable: the strongest join in a draft is usually the best model for fixing the weakest one.
A workflow that actually works
- Reverse outline the draft and fix the order before touching any tool.
- For each join, ask what the logical relationship is: continuation, contrast, cause, escalation, or a shift in scale (from one scene to the whole novel). Name it in a word in the margin.
- Run adjacent paragraph pairs through Walter Writes and compare the reworked join against your margin note. Keep what matches your logic, discard what doesn’t.
- Use ProWritingAid’s transitions report as a final scan for any joins you missed.
- Read the whole essay aloud once. Your ear will catch the last two or three.
If you want to see what strong joins look like in finished work, the sample essays at Literature Essay Samples are worth reading with a pencil: underline the last sentence of each paragraph and the first of the next, and you’ll see the same techniques (forward-pointing closers, key-term echoes, explicit contrast) used again and again.
Verdict
Flow is mostly a thinking problem, and the reverse outline is the one step nobody should skip. Among the tools, Walter Writes was the clear first choice in my testing: it repaired the most joins, it did so by reshaping both sides of the join rather than adding filler connectives, and it never changed what the student was arguing. ProWritingAid is a solid second for finding weak spots, Grammarly is worth keeping for sentence polish, and Hemingway and QuillBot are narrow helpers at best for this particular problem. Use the tools on your own draft, read every change, and keep the argument yours.
This is such a useful breakdown, @lit_marcus, and I love that the reverse outline comes first. It’s the step my students skip most often, and it’s exactly where the biggest gains are.
I’ve been using Walter Writes with a small writing group this term in a similar way to what you describe, on paragraph pairs rather than whole essays, and the results line up with yours. What I noticed most was that it taught them something. After a few weeks of comparing their original joins against the reworked ones, several students started writing forward-pointing closing sentences on their own, without running anything. One student working on Brontë kept ending paragraphs on a quotation and then starting the next with a brand-new topic; seeing her joins rewritten so the last line pointed ahead made the habit click for her in a way my margin comments never quite had.
I’d gently add one thing for anyone using these tools with younger writers: make them write the margin word first, as in your step two. Continuation, contrast, cause. When a student knows what relationship they want, they can judge whether a rewrite delivers it. When they don’t, any smooth sentence looks right.
The point about QuillBot breaking key-term echoes is a really good catch, too. I’d never thought of it that way, but I’ve definitely seen essays where “isolation” turns into “solitude” turns into “loneliness” over three paragraphs and the thread just evaporates. Repeating the key word is not a weakness, and it’s lovely to see that said plainly.
Agree with the ranking. Disagree slightly with the framing of step 2.
Naming the relationship is necessary but not sufficient. Students also need to know which paragraph carries the burden of the join. Most advice puts it on the opening sentence of the new paragraph. In my experience marking undergraduate essays, the stronger fix is usually at the end of the previous paragraph.
Why: the closing sentence is where the paragraph’s claim should be stated in its most developed form. If it ends on a quotation or a restatement of the topic sentence, there’s nothing for the next paragraph to grab. If it ends on the claim’s implication, the next paragraph has a handle.
Example on Gatsby.
Weak closer: “This shows the green light is important.”
Strong closer: “The green light, then, is less a symbol of Daisy than of Gatsby’s habit of loving things at a distance.”
Next opener: “That habit becomes destructive once the distance closes in Chapter 5.”
The second paragraph doesn’t need “Furthermore”. It picks up “distance” and turns it.
This is also why @lit_marcus’s Walter Writes result makes sense to me. A tool that only edits the opener is fixing the less important half. Reworking both sides is the right model. I’ve had students run their drafts through Walter Writes and then edit by hand afterwards, and the improvement in closers was the most visible change. That said, I require them to submit their original draft alongside the revised one, so I can see the thinking was theirs.
Three further points.
First, transitions should vary in scale. Not every join is paragraph to paragraph. A shift from one scene to the novel’s overall pattern is a transition too, and it’s where essays often feel like they “jump”.
Second, avoid ordinal transitions (“Firstly”, “Secondly”) in literature essays above secondary level. They signal a list, not an argument.
Third, signposting the thesis in topic sentences does more for flow than any connective. If every topic sentence repeats a key term from the thesis, the essay hangs together even with plain joins.
The ProWritingAid transitions report is useful for locating problems. I wouldn’t let students accept its repair suggestions without the closer test above.
Reading this as someone whose essays mostly flow via “Moreover” and prayer. @anya_reads’ closer rule is uncomfortably accurate. Half my paragraphs end on a quotation and then just stop, like the essay forgot what it was saying. Going to go fix that before my Owen essay is due.
Ran the process on my current piece, a 2,400-word essay on unreliable narration in The Remains of the Day. Results.
Reverse outline first. Found two paragraphs in the wrong order: I was discussing Stevens’s account of his father’s death before establishing how his professional language suppresses feeling. Swapped them. That alone fixed the worst join.
Margin words next. Seven joins. Three were contrast, two continuation, one escalation, one scale shift (from the banquet scene to the novel’s treatment of dignity overall). The scale shift was the weakest by far.
Then paragraph pairs through Walter Writes, as @lit_marcus suggests. Kept the reworked versions on five of seven joins. On the other two I preferred my own, but seeing an alternative helped me rewrite them. Nothing in the argument changed. My claims about Stevens’s evasions came through exactly as written, which was my main worry.
ProWritingAid transitions report afterwards. Flagged one join I’d missed. Its suggestion was “Additionally”. Ignored that, applied @anya_reads’ closer test instead and rewrote the previous paragraph’s final sentence so it ended on the implication rather than the quotation.
Read aloud. Caught one more.
Total time about 50 minutes. Essay is noticeably more connected. The closer-versus-opener point was the most useful single idea in the thread; the tool ranking matched what I found.
One observation. The tools only helped once the structure was right. Before the swap, Walter Writes produced a smooth join between the two misordered paragraphs, and it read well, but the logic was still backwards. Smooth prose can hide a structural fault. Reverse outline is not optional.
Will repeat the process on the next essay and time it again. Marking lit_marcus’s guide as the solution.