One reputation number, every month.
The tracker follows everything your organisation announces, finds how the press and AI assistants cover it, and turns that into the Media & AI Reputation Score: one number you can take to the board, with every article behind it one click away.
The Media & AI Reputation Score
A score from 0 to 100 for each calendar month, built from two parts.
- Press
- The average sentiment score of every article published that month that was read and scored. An article that couldn't be read is not coverage and isn't counted.
- AI answers
- The average tone of the AI assistants' answers that month: positive counts as 75, neutral 50, negative 25.
- The score
- The average of the two parts, rounded. 50 is neutral. Your first scored month is the baseline every later month is compared with.
A switch above the score shows it three ways: all coverage (the board number), professional press only, or citizen media only (bloggers, fan sites, newsletters and supporters' groups on your lists). Sites not on your lists count as professional press. AI answers count in every view.
Below the stories, the tracker also shows how your statements move coverage: for each story with coverage on both sides of the announcement, the change in average tone from before it to after it, and the average tone of articles that quote your organisation compared with those that don't.
The evidence is shown with the number
- What it rests on
- Directly under the score: how many scored articles it's based on, from how many outlets, and how many AI answers. A month with fewer than 15 scored articles from 5 outlets carries an Indicative label.
- Three lines, not one
- The chart draws the score with press tone and AI answers tone as their own lines, so you can see which one moved it. Bars underneath show how many articles were scored each month.
- What moved it
- The coverage pulling the score down and the coverage lifting it, up to three articles each, with the outlet, the date, the reason the AI gave and a link to the story.
- What's counted
- A line beside the coverage switch says how many scored articles are included, since which month, and from how many of your stories.
Step by step
What happens from announcement to score.
- 1
Your announcements become stories
Throughline reads your organisation's own news page. Each new post is screened by AI: real announcements (strategy, governance, finances, appointments, partnerships, squad and coaching news, statements) become stories to track. Match reports, previews and galleries are skipped. As soon as the news page has been read you're told how many new posts there are, how many were skipped and which ones the AI is reading; each of those then shows as added or not an announcement.
Press Check for new announcementsSee what's being checked straight awayKey messages pulled out of each one - 2
Monitor finds the coverage
Press Check for new coverage and Throughline runs Monitor on every story from the last three weeks; you can also press Monitor on a single story. For each one it looks for coverage published from a week before the announcement to three weeks after. It reads the latest posts from the people and outlets on your Who we watch list first, then searches the web, checking your outlets before looking more widely.
Your followed people's feedsYour outletsThe wider web - 3
Every article is read and checked
Each find is opened and read in full. An AI check confirms it's really about this story; anything off-topic or unreadable is rejected and listed with the reason, so nothing is counted silently. An article that turns up under two versions of its link is counted once. Up to eight new finds are checked per run.
Read in fullRelevance checkEach article counted once - 4
Each article is scored
Every article gets a sentiment score from 0 to 100 (50 is neutral), a tone in words, and a fidelity rating for how closely it carries your key messages. Quotes from your organisation are pinpointed, so you can see what was said and how it was reported.
Sentiment 0–100Tone and fidelityYour quotes pinpointed - 5
AI assistants are asked too
When you press Monitor on a current story (announced in the last three weeks), and at most once a month per story, Throughline asks one question per key message of ChatGPT, at OpenAI with its own web search, and of Gemini, with Google search. It records each answer word for word, its sources and its tone. Every question and answer is kept on the story page.
One question per key messageAnswers, sources and toneAt most monthly, when you press Monitor - 6
Watch a live story
For a story that's still moving, press Watch. Throughline re-checks it for new coverage every 30 minutes while its tracker or story page is open in a browser. Watching doesn't ask the AI assistants again.
Re-checks every 30 minutesCoverage onlyWhile the page is open
Who we watch
One tab on each tracker holds the outlets and people Throughline keeps watch on for you, and you can edit it. Outlets are the media that matter to you; they are searched first, and clubs and official bodies start switched off because they aren't independent media. Journalists and creators are the people who cover you, from staff correspondents to bloggers, newsletter writers and supporters' groups. Where they publish a written feed (an author page, blog or newsletter) and are switched on, their new posts are read on every Monitor run. Podcasts, YouTube and Bluesky are listed but not read.
The people are your agency's contacts: the same list your campaigns' pitches use, so nobody is entered twice. Open a name to see everything they wrote in your coverage, with its tone, how closely it carried your messages, and an overall score.
Who covers you
Who we watch is the list you choose; Who covers you is what actually happened. This tab groups every scored article three ways: journalists (named writers at professional outlets, from the byline), citizen media (the blogs, newsletters and fan sites you list as citizen media) and media outlets (every professional outlet, signed pieces or not). Each row shows how many pieces, their tone toward you, how closely they used your wording, how often they quoted your spokesperson, which stories they covered and the pieces themselves. The same section is in the report you share with your client. A writer who isn't one of your contacts yet has an Add to contacts button.
Looking back at past months
The Past months panel on the Stories tab fills in history from before you started, in two steps you run yourself.
- Bring in past announcements: choose a start date and Throughline reads the organisation's news page back to it, screening each post the same way as new ones.
- Find coverage for past stories: one month at a time, Throughline searches the web and your followed bloggers' and newsletters' archives for what was published around each announcement, then reads and scores it.
- Past months are press tone only. AI assistants can only say what they think today, so they aren't asked about past stories, and a past month isn't like-for-like with a month that includes AI answers.
- Older coverage is harder to find, so earlier months usually have fewer articles. The shared report and the PDF show how many articles are behind each month.
What it doesn't measure
- It measures how the press and AI assistants describe you, not what the public thinks. It isn't a poll.
- TV, radio, podcasts, YouTube and social media aren't scored. Podcasts, YouTube and Bluesky accounts can be listed so you can see who covers you.
- Coverage is found by reading followed feeds and by AI web search. Neither is guaranteed to catch everything, so the method and every rejected result are shown on each story.
- Coverage in Welsh and other languages is read and scored against your key messages. Welsh has been checked on a handful of articles so far, not tested at scale.
- Scores are worked out from the current records, so a month can move if more of its coverage is found later.
- The score only reflects the stories you've pressed Monitor on. The tracker says how many that is, next to the coverage switch.
- A month with few articles is labelled Indicative: one article can move its score by several points, so read it as a signal, not a result.
Everything behind the number is shown
Each story page lists the latest search terms used, the dates covered, every article with its score, every rejected result with the reason, and every AI question and answer.