Troubleshoot a Keyword Combiner by auditing the normalized input lists, theoretical A × B pair count, separator, order, and output-collision summary before copying any phrases. The Keyword Combiner creates every pair between two one-per-line lists and returns a deterministic, one-phrase-per-line result, so unexpected output usually points to input normalization, a pair-count limit, a separator choice, or a duplicate/collision rather than a hidden ranking problem. Its operating boundary is precise: one request may contain no more than 10,000 input pairs, checked before output deduplication. Exceeding that threshold fails the whole request instead of returning a partial sample. Within the limit, non-empty lines are trimmed at their outer edges, blank lines are ignored, exact case-sensitive duplicates within a source list are removed while preserving the first occurrence, and later duplicate final phrases are also removed. Because the combination logic runs in the current browser tab, you can reproduce a run, inspect its summary, and correct the source data without uploading either list to the widget.
Start by deciding whether the problem is a rejected request, a count mismatch, a formatting surprise, or a poor-fit keyword set. Each category requires a different check: limits for a rejected request, summary metrics for counts, exact settings for formatting, and manual research review for keyword quality. Do not begin with a near-match replacement; the fastest diagnosis is to compare the normalized inputs and displayed totals with what you intended to submit.
Match the symptom to its cause
Use this triage table before editing either list. It separates tool constraints from formatting assumptions and from the separate task of judging keyword quality.
| Symptom | Likely area | First check |
|---|---|---|
| No output or a rejected request | Input or pair constraint | Check term validity, editor and list limits, then the theoretical A × B count. |
| Fewer output lines than expected | Source duplicates or final collisions | Read the removed input duplicate and removed output collision counts. |
| More output lines than expected | Normalized source sizes or separator | Compare the displayed list sizes and confirm the exact separator. |
| Unexpected spacing or capitalization | Formatting assumption | Check internal spaces and remember that capitalization is not normalized. |
| Combinations sound unnatural or irrelevant | Keyword quality review | Review relevance, intent, grammar, landing-page fit, and demand outside the combiner. |
If the summary and source lists match your intention, the combiner has probably done its job. Any remaining suitability problem must be reviewed as a content or keyword decision rather than attributed to the combination engine.
Run a complete Keyword Combiner check
- Prepare both source lists. Paste one base term per line into list A and one modifier or second term per line into list B. Each line can contain a multiword phrase. Outer whitespace is trimmed, blank lines are ignored, and exact duplicate lines are removed after trimming while preserving their first occurrence. Internal spaces, punctuation, accents, emoji, and non-Latin scripts remain unchanged.
- Set the order and separator. Choose A-then-B or B-then-A, then enter the exact literal separator. It can be empty, a space, a hyphen, a plus sign, or another short sequence. No hidden spacing is added. The selected order determines the outer and inner list order, so changing it also changes the sequence of the generated phrases.
- Check the normalized sizes and theoretical pair count. Read the displayed list A and list B totals and calculate A × B. The pair budget is calculated after source-list deduplication and before output deduplication. A request accepts at most 10,000 input pairs; the next pair fails the whole request before output is generated.
- Combine the complete pair set. Run the combination once the inputs and settings are correct. A successful request generates the complete Cartesian product in deterministic first-seen order. It does not return a sampled list, a partial batch, or the first portion of an oversized result.
- Audit the summary. Review the unique output count, normalized list sizes, theoretical input-pair count, removed input duplicates, and removed output collisions. Narrow the source lists if the output is unexpectedly large. If it is unexpectedly small, check for exact repeated terms or different pairs that produce the same final phrase.
- Review candidates before copying. Remove irrelevant candidates in your review workflow. If a poor candidate comes from a weak base term or modifier, correct that source line and rerun the combination. After approval, press the copy button to copy the deterministic output as a single block with one phrase per line.
Audit pair counts, source duplicates, and output collisions
The theoretical input-pair count is the normalized list A size multiplied by the normalized list B size. For example, three colors and two product names give 3 × 2 = 6 input pairs; with no source duplicates or output collisions, the unique output count is also six.
| Constraint | Exact rule | Troubleshooting action |
|---|---|---|
| Each source term | At most 100 Unicode code points; control and Unicode line or paragraph separator characters are rejected. | Shorten long terms and remove prohibited characters. |
| Each source editor | Up to 100,000 UTF-16 code units. | Reduce pasted text and remove unnecessary lines. |
| Each source list | Up to 500 unique, case-sensitive terms. | Remove exact repeats and irrelevant entries. |
| Separator | Up to 20 Unicode code points; an empty value is allowed, but line breaks and control characters are not. | Shorten the separator or remove prohibited characters. |
| Input-pair request | At most 10,000 pairs, calculated before output deduplication; the next pair fails the entire request. | Narrow one or both source lists before combining. |
Source duplicates and output collisions are separate issues. Source deduplication is exact and case-sensitive after trimming, so “Shoe” and “shoe” remain distinct and may represent intentional variants. Repeated source entries lower the normalized list size and therefore the theoretical pair count. The summary reports how many such input duplicates were removed.
An output collision occurs when different input pairs create the same final string. With an empty separator, “ab” combined with “c” and “a” combined with “bc” both produce “abc.” The tool keeps the first final phrase and reports the later one as an output collision. For a focused cleanup process, see how to fix duplicate keyword combinations in a Keyword Combiner.
Verify order and separator behavior
The selected order determines which list is the outer loop and which is the inner loop. It changes both the loop sequence and the order of the resulting phrases. The output follows first-seen source order rather than alphabetical order, search volume, or an estimated ranking. When comparing two runs, keep the selected order and separator unchanged while changing the list contents, so that any difference is attributable to your list edits.
A separator is inserted literally, with no automatic spacing. With one normal space, “red” and “shoes” become “red shoes”; with an empty separator, they become “redshoes.” Empty separators also make final collisions more likely, so check the collision count when using one. For a detailed comparison of literal separator choices, see how to choose a separator in a Keyword Combiner.
Confirm complete output and local processing
A successful run generates every input pair, applies final collision handling, and returns the complete result in first-seen order. It never paginates, samples, silently cuts off, or replaces omitted content with an ellipsis. If the request exceeds the input-pair budget, the whole request fails before output generation, so there is no partial batch that could be mistaken for a complete list.
All combination logic runs in the current browser tab. The lists are not uploaded or stored by the widget, and copying occurs only after you press the copy button. Identical inputs, order, and separator produce identical text, making reruns suitable for version comparison and deterministic checks. The copied block can then be reviewed for a spreadsheet, ad-group draft, content brief, internal taxonomy, or later analysis in another tool.
Review candidates before using them
Treat every generated phrase as a candidate, not an automatically worthwhile keyword. The combiner does not estimate search volume, competition, cost per click, intent, language, relevance, or grammatical quality. It does not query a search engine, advertising account, keyword-data provider, or AI service. Some combinations will sound unnatural, overlap in meaning, or have no measurable demand, so validate important candidates with appropriate first-party or paid keyword evidence.
Before using the list in a campaign or publishing workflow, remove irrelevant phrases, confirm landing-page fit, review policy and trademark concerns, and validate demand. The tool also does not create advertising match-type syntax, quote phrases, add brackets, normalize capitalization, expand synonyms, stem words, or remove trademarked terms. Apply the rules of the destination system yourself.
The most reliable troubleshooting loop is to inspect the summary, correct the source lists, verify the order and separator, rerun the combination, and then review the candidates. Because the process is deterministic and local, changes should appear as explainable count and formatting differences rather than unexplained output.