📗 This content is the AI Max Analyzer user manual, the same one the script writes in the «Manual» tab of its spreadsheet. Here you can search it, listen to it, highlight it and save it to favourites.
To the right of the Top Search Terms table (from column O onwards) there are two analysis blocks. They are NOT automatic recommendations — they are statistical summaries over the full list of search terms in the period. The decision on what to do is always yours.
How it is built: The script breaks each search term into individual words (tokens). It strips accents (móvil = movil), lower-cases, filters stopwords ("de", "el", "the", "and"...) and drops words shorter than 2 characters or made of digits only. The same word counts once per search term even if it is repeated inside the query.
ACCOUNT-level analysis: IMPORTANT: the analysis is PER WORD, not per search term. For each word we aggregate the metrics of ALL the search terms in the period where it appears (regardless of whether each individual term converted). Then we classify each word into ONE block based on the aggregate TOTAL. A word NEVER appears in both blocks at once — it is either green or red depending on the TOTAL of accumulated conversions.
★ block (green): "WORDS IN TERMS THAT CONVERTED". Words whose sum of conversions (across all the search terms where they appear) is > 0. Sorted by accumulated conversions descending. Columns: Word, # Terms, # Camp., Cum. cost, Cum. conv., Cum. CPA (= Cost / Conv. aggregated), Cum. impr., Cum. clicks. It tells you which keywords are contributing to conversions.
✗ block (red): "WORDS IN TERMS THAT DID NOT CONVERT". Words whose sum of conversions is exactly 0 — 100% waste in the period. Sorted by accumulated cost descending (the ones with most cost and no return first). Columns: Word, # Terms, # Camp., Cum. cost, Cum. impr., Cum. clicks. They are the clearest candidates for a broad negative.
Shared columns: "# Terms" = in how many different search terms that word appears. "# Camp." = in how many different AI Max campaigns it appears (it is the exact number of campaigns that will be modified when you tick the "Broad negative" checkbox). The metrics are the SUM of every search term containing the word.
"Broad negative" and "All campaigns" checkboxes: Every word has TWO checkboxes at the start of its row, and EACH ONE fires on its own (you do not need to tick both): "Broad negative" → adds the word as a BROAD MATCH negative only in the campaigns where it appears (the ones counted by "# Camp."); "All campaigns" → adds it across ALL your AI Max campaigns in the account. If you tick both, "All campaigns" wins. If you tick neither, nothing happens. Broad match blocks EVERY search term containing that word in any order — more aggressive than exact match. A note on accents: the analysis groups "aplicación" and "aplicacion" as a single word, but the script creates one negative per real spelling (with and without the accent) because Google does not block accented variants automatically. Every negative added is logged in "Negatives Added" with match type "Broad".
Careful with the green block: If you tick a word from the green block as a broad negative, bear in mind that word is contributing to your overall conversions. Adding it as broad will block ALL the search terms containing it — including the ones producing those conversions. The script does not stop you (there can be legitimate cases: your own brand, terms with an out-of-control CPA even with conversions, etc.) but it leaves a "⚠ Converts globally (X conv.)" note in the history for traceability.
Granularity and volume: The green block shows up to 30 words; the red one up to 50 (wider, because a negative decision needs more options). If a word appears in a single search term, it does NOT show in either block — repetition is the statistical signal we are after.